(PDF) Voting Advice Applications in Europe: The State of the Art

VOTING ADVICE APPLICATIONS

IN EUROPE

THE STATE OF THE ART

Edited by

LORELLA CEDRONI & DIEGO GARZIA

Napoli, 2010

Proprietà Letteraria Riservata

CIVIS s.n.c/Scriptaweb – Napoli

È vietata la riproduzione di questo libro o parte di esso con qualsiasi

mezzo tecnico.

Table of Contents

Preface .............................................................................................. 9

Chapter One

The Effects of VAAs on Users’ Voting Behaviour: An

Overview

Diego Garzia .................................................................................... 13

Chapter Two

The Irresistible Rise of Stemwijzer

Jochum de Graaf ............................................................................... 35

Chapter Three

Much Ado About Nothing? Online Voting Advice

Applications in Finland

Outi Ruusuvirta ............................................................................... 47

Chapter Four

The Impact of Voting Indicators: The Case of the German

Wahl-O-Mat

Stefan Marschall, Christian K. Schmidt ............................................ 65

Chapter Five

More than toys? A first assessment of voting advice

applications in Switzerland

Andreas Ladner, Gabriela Felder, Jan Fivaz ................................... 91

Chapter Six

Do the Vote Test

: The Belgian Voting Aid Application

Michiel Nuytemans, Stefaan Walgrave, Kris Deschouwer ............... 125

Chapter Seven

Mapping the Political Landscape: A Vote Advice

Application in Portugal

Marina Costa Lobo, Maarten Vink, Marco Lisi ........................... 143

4

Chapter Eight

wahlkabine.at

: Promoting an Enlightened Understanding

of Politics

Christine Mayer, Martin Wassermair ............................................. 173

Chapter Nine

cabina-elettorale.it

(Provides advice to Italian voters since

2009)

Roberto De Rosa ............................................................................. 187

Chapter Ten

Are the Voting Advice Applications (VAAs) Telling the

Truth? Measuring VAAs' Quality. Case Study from the

Czech Republic.

Michal Škop ................................................................................... 199

Chapter Eleven

The use of Voter Advice Application in Poland –

Glosuje.com.pl

Agata Dziewulska .......................................................................... 217

Chapter twelve

Voting Advice Applications in Europe: A Comparison

Lorella Cedroni ............................................................................... 247

REFERENCES LIST ........................................................................ 259

Notes on Contributors

Lorella Cedroni is professor of Political Philosophy at the

University of Rome “SAPIENZA” (Italy); PhD in Social and

Political Sciences, EUI (Florence). She was Fulbright Distin-

guished Professor at the University of Pittsburgh (PA), USA, in

2008. She is Director of the SEC (European Society of Cul-

ture) International Summer School, Venice, and coordinator

of the national research on VVA in Italy. Her research fo-

cuses on political representation, democracy and political

parties.

Kris Deschouwer is professor of Political Science at the Vrije

Universiteit Brussel. His research focuses on political parties

and party systems, with special attention to the role of par-

ties in complex and divided societies.

Roberto De Rosa is assistant professor of Public Policies at

the University of Viterbo (Italy). He obtained his PhD in

Science of Communication and Complex Organizations at

the University of Rome (Lumsa). His research is focused on

social capital and political parties. He collaborated with

cabina-elettorale.it.

Agata Dziewulska is assistant professor in international rela-

tions and European studies at the Centre for Europe, Uni-

versity of Warsaw. She holds a PhD in Social and Political

Sciences from the European University Institute in Florence

and specialises in political systems of post-conflict states.

She coordinates the academic part of the VAA Glosuje.com.pl

project in Poland.

Gabriela Felder is political scientist and member of Poli-

tools, a political research network owning the Swiss voting

advice application smartvote. She was the project coordinator

of the Swiss contribution to the EU-Profiler, a voting advice

6

application set up for the first time for the European Par-

liamentary Elections in 2009.

Jan Fivaz studied history, political science and economics at

the University of Berne. At the moment he is working on

his PhD on the quality of political representation in Switzer-

land at the Swiss Graduate School of Public Administration

(IDHEAP) at the University of Lausanne. He is also a

founding member of Politools – the owner of the Swiss

VAA smartvote.

Diego Garzia is PhD Candidate in Comparative and Euro-

pean Politics at the University of Siena. Previously, he stud-

ied at the University of Rome (Sapienza) and Leiden Uni-

versity. He has been editor-in-chief of cabina-elettorale.it and

co-editor of VoteMatch.eu. His research is focused on the

role of personality in orienting political attitudes and behav-

iour. He is also interested in parties, elections, VAAs, and

their impact on voting behaviour.

Jochum de Graaf is project leader of Stemwijzer at the Insti-

tuut Publiek en Politiek (IPP).

Andreas Ladner is professor for political institutions and

Swiss public administration at the autonomous university

institute IDHEAP in Lausanne. His areas of research in-

clude political parties, municipalities, institutional change

and e-democracy. He has conducted several major research

projects of the Swiss National Science Foundation. Actually

he leads a research project on the voting assistance applica-

tion (VAA) smartvote.

Marco Lisi holds a BA in Political Science and International

Relations (University of Florence) and MA in Political Sci-

ence (ISCTE, Lisbon). He obtained his PhD in Political Sci-

ence from the University of Florence in 2007. Research in-

terests focus on political parties, electoral behaviour, de-

mocratic theory and political communication.

7

Marina Costa Lobo obtained her Doctorate from Oxford

University with a thesis researching Prime Ministerial power

in Portugal. She is currently a research fellow at the Institute

of Social Sciences in Portugal, and is also a co-editor of the

Journal South European Society and Politics. Her interests

centre on electoral behaviour and political institutions in

Portugal in a comparative perspective.

Stefan Marschall is professor of political science at the Uni-

versity of Siegen. Before, he worked as senior lecturer in the

social science department of the University of Düsseldorf.

He is a specialist on the German political system and com-

parative as well as transnational parliamentarism. Since 2003

he is in charge of the Wahl-O-Mat research commissioned by

the Federal Agency of Civic Education.

Christine Mayer holds a degree in history and English lan-

guage and literature. She works at the Institute for New Cul-

ture Technologies/t0 and is responsible for the project

wahlkabine.at.

Michiel Nuytemans is a researcher at the University of

Antwerp and currently at work for Indigov, a spin-off com-

pany of the K.U.Leuven. His research interests are voting

aid applications, media and politics, elections and eParticipa-

tion.

Outi Ruusuvirta is PhD candidate in the Department of

Government at the London School of Economics and Po-

litical Science. Her doctoral research uses experimental

methods to investigate the effects of online voting advice

application use on citizens’ electoral behaviour in Ireland,

Finland and the United Kingdom.

Christian K. Schmidt is a lecturer of Political Science at the

social science department of the University of Düsseldorf.

His research focuses on political communication, voting

behaviour, political participation and political corruption.

8

He consults the Federal Agency for Civic Education in

Germany (bpb).

Michal Skop received Master in Mathematics and PhD in

Demography at Charles University in Prague. He is a co-

author of Czech and Slovak VAAs and watchdog website

KohoVolit.eu. He is interested in scientific approaches to

VAAs and in statistical modelling of roll-call data. He works

as a programmer nowadays.

Maarten P. Vink is a Post-Doctoral Research Fellow at the

Institute of Social Sciences of the University of Lisbon, Por-

tugal and Assistant Professor at the Department of Political

Science of the University of Maastricht, the Netherlands. He

holds a PhD in Political Science from the University of Lei-

den (2003). Vink's research interests are European integra-

tion, immigration, citizenship and comparative methodol-

ogy.

Stefaan Walgrave is professor in political science at the Uni-

versity of Antwerp (Belgium). He leads the Media, Move-

ments, and Politics (M²P) research group in Antwerp. His

research interests are media and politics, social movements,

electoral behaviour and campaigns.

Martin Wassermair is a historian and political scientist. He

works at the Institute for New Culture Technologies/t0 in

Austria, and is responsible for the project wahlkabine.at.

PREFACE

As the rise of the internet has transformed the world of in-

formation in democratic societies, so it has changed the pat-

terns of political communication in Western democracies.

With regard to electoral campaigns, the internet has quickly

established itself as an essential medium alongside the tradi-

tional media (newspapers, television). Contemporary publics

can choose from a growing number of partisan websites pro-

viding detailed information on the behalf of specific parties

and candidates. At the same time, the interactive possibilities

of the internet have generated new opportunities for accessing

non-partisan sources of political information. A major innova-

tion in this respect is represented by the so-called Voting Ad-

vice Applications (VAAs).

VAAs have literally taken Europe by storm in the past dec-

ade, with millions of voters turning to these web-based tests

at election time. In essence, VAAs are online databases of par-

ties’ positions on a number of core policy issues. By compar-

ing users’ position on those issues with that of the parties, the

application produces a sort of voting advice – usually in the

form of a rank-ordered list, at the top of which stands the

party closest to the user’s policy preferences. Such applica-

tions are aimed primarily at increasing voters’ understanding

of what parties stand for at a very cheap cost (in terms of pro-

curement, analysis, and evaluation) by means of an immediate

and enjoyable approach. In turn, this can increase voters’ in-

terest in political matters, motivate them to discuss about poli-

tics, and hopefully gather further information.

More recently, the impressive numbers of users visiting

VAA-websites have led some political scientists to hypothe-

size an effect of these tools also on voters’ electoral behav-

iour. Questions about VAAs’ ability to motivate users to turn

10

Preface

out and vote (possibly for the party advised by the applica-

tion) have been repeatedly raised. However, little empirical

evidence has been collected so far – this lack being particu-

larly evident in comparative perspective.

To overcome this gap in the literature, we decided to devote

our efforts to this first systematic investigation of the major

country-based VAAs from all over Europe. The cases in-

cluded range from established applications such as Dutch

StemWijzer and German Wahl-O-Mat to ‘first attempts’ from

Southern and Eastern Europe, in order to portray in detail the

various stages of development of the VAA-phenomenon

around the continent.

The time-point under analysis is, in each case, the European

election of June 2009. The second-order nature of this electoral

competition – where voters are thought to vote for parties

closer to their preferences and ideological outlook – seems in

fact an ideal context to assess the potential of such applica-

tions.

Clearly, the comparability of the empirical findings collected

throughout the volume is limited by the specificities of the na-

tional setting in which the various VAA operate, as well as by

the widely different methodology employed (and popularity

enjoyed) by such applications. For this reason, we look for-

ward to the huge amount of data collected by two ambitious

pan-European VAAs developed in occasion of the 2009 EP

election – EU Profiler and VoteMatch Europe1 – in order to as-

sess if and to what extent the electoral effects of country-

based VAAs are comparable with those exerted by these

Europe-wide experimental applications.

1 Although EU Profiler [www.euprofiler.eu] as such is not featured in

this volume, the methodology is widely discussed with reference to its

Portuguese spin-off Bússola Eleitoral (see Chapter 7). The same goes for

VoteMatch Europe [www.votematch.eu] which is a licensed version of

Dutch StemWijzer (see Chapters 1 and 2).

11

Preface

The modest hope of the present authors is that of having

ignited, through this volume, a fruitful exchange between aca-

demic and practitioners on applications that are likely to be-

come an ever more used and appreciated feature of electoral

campaigns in the years to come.

Lorella Cedroni & Diego Garzia

Rome & Siena – June 2010

CHAPTER ONE

The Effects of VAAs on Users’ Voting Behaviour:

An Overview

DIEGO GARZIA

Voting Advice Applications (VAAs)1 are becoming a wide-

spread feature of electoral campaigns in Europe, thus at-

tracting a growing interest from journalists, commentators,

and – more recently – political scientists. VAAs help users

casting a vote by comparing their policy preferences on ma-

jor issues with the programmatic stands of political parties

on such issues. The mechanism employed is rather straight-

forward: the respondents fill in a web-questionnaire with

their opinion on a wide range of policies; after comparing

the user’s profile with that of each party, the application

produces a sort of advice under the form of a rank-ordered

list, at the top of which stands the party closest to the user’s

policy preferences. To get a raw picture of the spread of

VAAs around the continent, Walgrave et al. (2008a) sur-

veyed a large sample of European political scientists: as

they find out, in 2007 there was (at least) one voting advi-

1 These applications have been defined in several ways: Voting Indicators

(Boogers and Voerman, 2003), Party Profile Websites (Hooghe and Teepe,

2007), Political Internet Consultants (Kleinnijenhius and van Hoof, 2008),

Online Vote Selectors (Ruusuvirta and Rosema, 2009) just to mention

some examples from the most recent literature. However, in this chap-

ter we will refer to them as Voting Advice Applications (Walgrave et al.,

2008) because of the seemingly emerging consensus on such denomina-

tion – or, at least, on the “VAAs” acronym (see: Nuytemans et al. in

this volume).

14

Diego Garzia

sor running in 15 countries out of the 22 surveyed. The

numbers are impressive: to mention just a few, suffice to

say that in 2006 the Dutch StemWijzer counted some 4.7

million advices given (equal to roughly 40 percent of the

Dutch electorate), while three years later the German Wahl-

O-Mat reached 6.7 million users (12 percent of the eligible

voters in the country). The widespread diffusion (in terms

of countries) and popularity (in terms of users) of these

tools is obviously linked with the rise of internet. On the

one hand, technological developments made easier the

production of VAAs; on the other hand, the pervasive dif-

fusion of the medium rendered them accessible to a huge

number of potential users without serious effort. However,

the internet alone cannot possibly account for VAAs’ suc-

cess. Also structural political changes going on in Western

publics must be taken into consideration: in particular, the

erosion of cleavage-based voting (Franklin et al., 1992) and

partisan alignments (Dalton and Wattenberg, 2000) aug-

mented dramatically the number of floating, undecided

voters (Dalton et al., 2000) and consequently the demand

for guidance (or at least, advice) in the – not anymore sim-

ple – act of voting.

1.1. The decline of long-term determinants of voting behaviour

According to classic democratic theorists such as Mill,

Locke, and Tocqueville, the prerequisite for a good democ-

racy is that citizens/voters posses an acceptable level of po-

litical abilities (in terms of knowledge, understanding, and

interest in political matters). However, with the rise of sur-

vey research in the 1940s it appeared immediately the stun-

ning contrast between the classic image of a supercitizen and

the real nature of contemporary voters (Berelson et al.,

1954; Campbell et al., 1960). Voting behaviour research

15

Chapter One - The Effects of VAAs on Users’ Voting Behaviour

looked then at the ways in which citizens “manage the

complexities of politics and make reasonable deci-

sions…through the use of political cues and other decision-

making shortcuts” (Dalton and Wattenberg, 1993: 196).

Early studies focused on two strong providers of political

cues: social groups and partisan affiliations. The former

draws the attention of voters to the party (or parties) that

historically have supported the social group to which the

voter belongs. Hence a voter can base a decision between

competing parties on cues such as the endorsement of la-

bour unions, business associations, religious groups, and so

on. An even more powerful source of political cues is parti-

sanship. While the usefulness of social group cues is limited

to topics directly related to group interests, party identifica-

tion has broader applications. Parties are in fact central ac-

tors of democratic politics, so almost all political phenom-

ena can be evaluated within a partisan framework (Miller,

1976). The social-psychological view holds that party iden-

tifications act to filter individuals’ views of the political

world, providing them not only with a means for making

voting decisions, but also with a means for interpreting is-

sues and candidates. For this reason, some have come to

define partisanship as the ‘ultimate cost-saving device’

(Fiorina, 1990).

By the 1970s dramatic changes began to affect Western

societies, and in particular the stable social cleavages on

which they were based. There are several causes for the de-

cline of social classes as such (e.g., embourgoisement of the

working class, growth of the service sector and governmen-

tal employment, increased geographic mobility and urbani-

zation); in political terms, the major consequence has been

a sharply reduced ability of social cleavages to structure in-

dividual voting choice (Franklin et al., 1992). As it has been

argued, “social cues may still be a potent influence on vot-

16

Diego Garzia

ing behaviour for people who are integrated into traditional

class or religious networks…but today there are far fewer

people who fit within such clearly defined social categories”

(Dalton, 1996a: 331). In the same years, it also became clear

that voters were de-attaching from political parties. Russell

Dalton (1984; 2000) linked this pattern of partisan

dealignment to a process of cognitive mobilization among

Western publics due to social modernization (Inglehart,

1977), and in particular to rising levels of education and the

spread of television as a source of political information. In

the light of this, it could be previewed a decreasing func-

tional need of partisanship (Shively, 1979) inasmuch voters

were increasingly able to orient themselves in the complexi-

ties of politics thanks to stronger cognitive skills as well as

less costly information, as provided by the new medium.

The more visible consequences of such process is the

growth of so-called apartisans (e.g., voters not attached to

any particular party, but yet equipped with the necessary

skills to understand politics) that resulted in an increased

aggregate volatility at the macro-level, and a progressive de-

laying of vote decision’s timing at the micro-level (Dalton et

al., 2000). These changes in the sources of political and

electoral cues led to what some authors named the individu-

alization of politics. This has involved

“a shift away from a style of electoral decision-making

based on social group and/or party cues toward a more indi-

vidualized and inwardly oriented style of political choice. In-

stead of depending upon party elites and reference

groups…contemporary publics are more likely to base their

decision on policy preferences, performance judgments, or

candidate images” (Dalton 1996b: 346).

17

Chapter One - The Effects of VAAs on Users’ Voting Behaviour

1.2. Candidates, issues and other short-term determinants of voting be-

haviour

Candidates have come to matter increasingly in vote

choice for different reasons (for a review, see: McAllister,

2007). However, voting on the basis of personality has of-

ten being seen as ‘irrational’ (Converse, 1964; Page, 1978),

for the popular cynical view of candidates is that “they are

affectively packaged commodities devised by image makers

who manipulate the public’s perceptions by emphasizing

traits with special appeal to the voters” (Dalton and Wat-

tenberg, 1993: 208). The literature has focused extensively

on what kind of voters are more sensitive to candidates’

personalities. According to Pierce (1993), “[c]andidate traits

need not be related to politics, whereas parties, ideologies,

and issues are inherently political; thus, candidate traits re-

quire less sophistication to understand and incorporate into

the voting decision” (24).

As said, however, a growing part of contemporary elec-

torates is basing their vote choice on a rather different as-

pect, that is, issue preferences (Franklin et al., 1992; Dalton,

1996a). The standard model of rational decision-making

based on issues, as applied to the study of voting behav-

iour, is the spatial model developed by Anthony Downs

(1957). According to Downs, every policy can be placed on

a left-right continuum. Issue voting basically means that

people’s vote is determined by the proximity/distance on

the continuum between their own position on the issue and

that of the parties. In order to be meaningful, issue voting

requires that:

a) voters have a clear policy preference;

b) parties offer competing proposals over the same policy

(Nie et al., 1979);

18

Diego Garzia

c) voters are able to link their position on the policy to that of

one of the parties (Butler and Stokes, 1969).

Contrary to candidate-driven vote, the key aspect here is

sophistication – especially in terms of knowledge and un-

derstanding of political matters. In order to link their policy

preference to that of the parties, voters need not only to

have developed a preference, but also to have gathered a

sufficient amount of information about the parties’ position

on those policies. Unfortunately information is costly. Three

are in particular the costs involved in becoming sufficiently

informed over an issue:

i) procurement: the costs of gathering, selecting, and trans-

mitting data;

ii) analysis: the costs of undertaking a factual analysis of

data;

iii) evaluative: the costs of relating data or factual analysis to

specific goals (Carmines and Huckfeldt, 1996: 245).

Furthermore, things can get way more complicated be-

cause of the multi-party nature of many democratic systems

(thus information on each issue should be gathered with re-

spect to every competing party) and from the serious pos-

sibility that there is more than one relevant issue to the

voter (Pappi, 1996). If this is the case, then “an issue voter

would vote for the candidate who was closer to him on the

two-dimensional plane defined by the two issues” (Nie et

al., 1979: 159). Admittedly, this may be complicated.

19

Chapter One - The Effects of VAAs on Users’ Voting Behaviour

1.3. The Use of VAAs: An Advice for Issue Voters

Here is where VAAs come into the picture. Voting advice

applications help voters in making cheaper the costs in-

volved in getting informed, because:

i) the VAA-makers have already procured all the relevant in-

formation with respect to the parties’ positions on major

policy issues;

ii) the application analyzes the information through an

automated algorithm that compares the position of the

voter on such issues with those of the parties, and…

iii) …assists the voter in evaluating the information by pro-

viding a ‘vote advice’ – that is, a rank-order list at the

top of which stands the party closer to the voter on the

n-dimensional issue space (Edwards, 1998).

In this way, users can fulfil the requirements of meaning-

ful issue voting simply by having a clear preference over

policy issues. The application will do the rest, linking the

voter’s position on the various policies with that of the par-

ties (Butler and Stokes, 1969).

How does a VAA work2? There are several underlying as-

sumptions behind VAAs development: namely, that there

2 In this section we refer to the so-called ‘StemWijzer Method’ (for a

better discussion, see: de Graaf in this volume) as developed by the

Dutch Instituut voor Publiek en Politiek, on the grounds that it has been

the pioneer VAA in Europe (the first release dating 1989) and among

the most successful in terms of users. Furthermore, many European

applications are based on a fundamentally similar methodology (e.g.,

Austrian Wahlkabine, Belgian Doe De Stemtest!, Italian Cabina Elettorale),

while some are licensed versions of StemWijzer itself (e.g., Bulgarian

Glasovoditel, French Mon Vote à Moi, German Wahl-O-Mat). Of course,

some VAAs differ more substantially from StemWijzer: To their innova-

20

Diego Garzia

are substantive programmatic differences between the po-

litical parties, that many voters are willing to base their vot-

ing decision on an assessment of these differences but they

have difficulties in making sense of these differences (Ed-

wards, 1998). VAAs help voters overcome these difficulties

by lowering the costs related to the procurement, analysis

and evaluation of information, as described below.

Information procurement. The first job of the editors is to se-

lect key issues raised in the various parties’ electoral mani-

festos under the form of specific ‘statements’ (e.g., ‘taxation

should be reduced regardless of the taxpayer’s income’).

Suitable statements must be politically relevant, tackle a

number of diverse issues, and discriminate between parties3

(Walgrave et al., 2009: 1168). The final list of theses is given

to the recognized members of political parties (usually,

those who drafted the party manifesto), who decide

whether the party agrees, disagrees or is neutral to the given

statements.

Information analysis. The application gets online some

months before the Election Day, as soon as all the parties

have answered to the statements’ list and this has been re-

fined by the editors according to the selection criteria out-

lined above. The user is provided with a number of policy-

related statements, to each of which (s)he can choose to

agree, disagree, or stay neutral (see Figure 1.1). When all

statements are answered, the voter can also assign an extra-

weight to those issues (s)he feels of particular importance.

Then the application compares the answers of the user

with those provided by each party, on the basis of the prin-

tive features and the reasons underlying their implementation is de-

voted a discussion in the final section of this chapter.

3 In other words, issues on which all parties agree are ruled out from

the questionnaire. On these grounds valence issues (Stokes, 1963) are

never included in the making of VAAs.

21

Chapter One - The Effects of VAAs on Users’ Voting Behaviour

ciple of shortest distance (for a detailed description, see: de

Graaf in this volume).

Figure 1.1 – Example of a VAA statement

Source: <http://www.votematch.eu>

Information evaluation. Eventually, the user is provided with

a vote advice – as said, a rank-order list at the top of which

lies the party placed at the shortest distance to the user4 (or,

to put it more easily, the party whose answers to the ques-

tionnaire are most similar to those provided by the user). In

this way, the VAA helps the user evaluating the informa-

4 Hence, VAAs favor a proximity model of issue voting (Downs, 1957) ra-

ther than a directional one in which voters are assumed not to care so

much about the exact political distance, but rather prefer those ‘on

their side’ (Rabinowitz and MacDonald, 1989).

22

Diego Garzia

tion by showing which parties are closer (farther) to his/her

issue preferences (see Figure 1.2).

Figure 1.2 – Example of a ‘vote advice’

Source: <http://www.cabina-elettorale.it>

What has been said so far relates to the intentions of VAAs

developers; but what about the users? Are they playing the

tool for the same reasons (e.g., gather political informa-

tion)? A study of political websites’ users in the Nether-

lands by Boogers and Voerman (2003) demonstrates that

23

Chapter One - The Effects of VAAs on Users’ Voting Behaviour

the main reason for visiting a VAA website is exactly to

collect more information on the positions of political par-

ties. Furthermore, it has been showed that the majority

ofVAA users get strengthened in their political knowledge

after having played the test, while one in three declares to

feel more involved in politics (ibid.).

1.4. The impact of VAAs on political participation

VAAs can be thought to affect political participation in at

least three ways: (1) by contributing to a change in indi-

viduals’ information-seeking behaviour: that is to say, moti-

vating users to gather further information about politics

and political parties. We will refer to this item as the ‘cogni-

tive dimension’ of participation (see: Marschall and

Schmidt in this volume); (2) by motivating people to turn-

out, even if they had not intended to do so before playing

the tool. In this case, the supposed VAA-effect would quan-

titatively affect the ‘behavioural dimension’ of political par-

ticipation (e.g., to vote or not to vote); (3) finally – yet less

frequently (Ruusuvirta and Rosema, 2009) – by affecting

individuals’ vote intentions: that is, convincing already de-

cided voters to change their political preference. Here, the

VAA would affect the quality of participation (e.g., voting

for party B rather than party A). Let us analyze in turn the

(potential) effect of VAAs on each of these spheres in

more detail.

Cognitive effects. The question here is what kind of differ-

ence VAAs make to the information-seeking behaviour of

voters and, consequently, their level of information (Ed-

wards, 1998). Motivating users to gather more information

can be relevant, since the “search for more information can

serve to increase one’s competences in understanding poli-

tics by affecting the extent and quality of individuals’ politi-

24

Diego Garzia

cal activities” (Marschall and Schmidt in this volume). Pre-

vious research tells us that this is indeed the case: a survey

conducted in 2005 among German VAA users shows that

slightly more than a half of the respondents declare to be

motivated to collect further political information after hav-

ing performed the test (Marschall, 2005). Of course, cogni-

tive effects would be of little interest to our purposes if we

could not detect any reflection in the actual behaviour of

users/voters. However, previous research has shown that

VAAs can also affect vote choice, in both quantitative

(turnout) and qualitative (vote intention) terms.

Behavioural effects. As a result of long-term patterns of so-

cial and partisan dealignment, contemporary electorates

have developed an increasingly instrumental orientation to-

wards politics (Thomassen, 2005). Accordingly, voting has

become something that citizens will only do as long as they

have a real choice, and such instrumental orientation im-

plies that “voters will decide from election to election

whether they will vote or not” (ibid., 255). Consistently with

low-information rationality theories (Popkin, 1991), the prob-

ability for such instrumental voters to cast a vote is in-

versely proportional to the effort required to gather enough

information. Hence, we can hypothesize that the usage of

VAAs, by lowering the cost of information, facilitates vote

decision and thereby increases the chance of voting vis-à-vis

abstention. This hypothesis is supported by previous stud-

ies on the impact of political knowledge on turnout, which

show that to higher levels of political information corre-

sponds a decreased likelihood to abstain (Delli Carpini and

Keeter, 1996). Research on the impact of VAAs usage on

turnout further reinforces this conclusion. A study by Myk-

känen and Moring (2007) on Finland found that, after con-

trolling for demographic variables, using a vote selector in-

creases the likelihood of voting by 21 per cent for men and

25

Chapter One - The Effects of VAAs on Users’ Voting Behaviour

23 per cent for women in the 2003 parliamentary elections.

Lower, yet significant figures, are reported by Marschall

(2005) for Germany (8% of respondents ‘considered’ the

option of voting in the federal election of 2005 thanks to

the test) and Boogers (2004) for the Netherlands (12%

considering voting in the parliamentary election of 2003).

The reasoning above holds for voters whose preference is

weak (when not a real lack of preference) – hence perform-

ing the test might have given enough strength to a pre-

existing preference to get to the polls. But what about users

who faced the test with strong(er) preferences already

shaped in their mind? Based on what we know about cogni-

tive dissonance (Festinger, 1962), it is likely that a wide major-

ity of these voters will reject a recommendation which does

not conform to previously existing beliefs. This hypothesis

is strongly supported by empirical evidence; yet, the very

same studies show that there is indeed a correspondingly

small but significant proportion of VAAs users who declare

that they will change their vote choice in accordance with the

voting advice received. The proportion of self-declared

‘swing-voters’ among VAAs users seems, according to the

available evidence, quite varied across different national set-

tings: 3 percent in Finland (Mykkänen and Moring, 2007), 6

percent in Germany (Marschall, 2005), and up to ten 10

percent in the Netherlands (Kleinnijenhuis and van Hoof,

2008). Furthermore, one could compellingly ask whether

this reported ‘intention’ to change vote preference is

matched by actual changes in voting behaviour. To this is

devoted a study by Walgrave et al. (2008a) who compare the

results of a survey performed by Belgian users after playing

Doe De Stemtest! [in English: Do the Vote Test, DVT] with a

post-electoral one in which respondents were asked

whether their post-test intention persisted until the polling

day. According to the former, only “one in 10 users said

26

Diego Garzia

that DVT contributed to their doubt and barely one in a

hundred said that DVT made them change their mind”

(ibid.: 59). These figures gets significantly downsized in the

latter survey, which documents that

“on average, only half of the people who said DVT made

them doubt about their vote actually changed prefer-

ences…Even among the small group of people saying that

DVT really made them change their mind, one third did not

change their mind at all and remained loyal” (Walgrave et al.,

2008a: 65-66).

If the figures are correct, then it would seem that a sub-

stantial proportion (around two thirds) of users who in-

tended to change their vote choice after playing the test did

actually so at the ballot. Combining this figure with the (con-

tinuously growing) number of visitors to VAA websites

would give us a tentative – yet quite appealing – measure of

the potential of these tools to affect election outcomes. Let

us take the Dutch case for a small thought experiment:

StemWijzer counted 4.7 million users to its 2006 edition

(that is, 40 percent of the Dutch electorate); after the test,

as said, an abundant 10 percent of surveyed users reported

the intention to change their vote in accordance with the

advice received; assuming that two in three did so at the

ballot, then we would get to roughly 300.000 votes moved by

StemWijzer (!) corresponding to 3 percent of votes casted in

that election. Clearly, it is just a speculation made up by ap-

plying a country-based (e.g., Belgian) finding on the figures

relative to a different national setting (e.g., the Nether-

lands). Much more solid evidence is needed before we can

generalize the effective impact of VAAs beyond the post-

test survey respondents. Yet these figures, along with the

rising popularity of these tools throughout Europe, seem to

call the attention of political scientist on a topic that the

27

Chapter One - The Effects of VAAs on Users’ Voting Behaviour

discipline has been overlooking, with few exceptions, for

too long.

1.5. Critiques and improvements to the ‘StemWijzer Method’

In the light of their impact on (a growing number of) us-

ers, an academic interest has recently arisen with respect to

the consistency and reliability – to put it in a word, the qual-

ity – of the voting advice provided by such applications.

The main critiques targeted to what we have called the

‘StemWijzer Method’ deal with four different aspects: a) the

selection of the statements to be included in the test; b) the

different saliency that both voters and parties attach to dif-

ferent issues; c) the way in which it is established the posi-

tion of each party on the various issues; and d) the dimen-

sionality of the policy space in which voters and parties are

placed. With regard to the first aspect, Walgrave et al.

(2009) demonstrate that the specific selection of statements

to be included in the VAA questionnaire has a considerable

impact on the advice provided to the user. According issue

ownership theories (Budge and Farlie, 1983), it goes without

saying that the inclusion of certain issues may result in an

advantage for those who own them; hence, the need for a

systematic, empirically-oriented process of statement selec-

tion (such as the one presented by Nuytemans et al. in this

volume). As to the second aspect, theories on issue salience

(Niemi and Bartels, 1985) have already highlighted that dif-

ferent voters and parties may (and do) consider some issues

more relevant than others. Since its very first version,

StemWijzer provided its users with the possibility to assign

an ‘extra-weight’ to some issues considered particularly

relevant. However, the same possibility was (and is still) not

provided to political parties. Some criticized this feature on

the ground that dichotomous answer levels (e.g., the party

28

Diego Garzia

agrees, the party does not agree) can reduce sharply the dis-

criminatory power of the tool (Krouwel and Fiers, 2008).

For this reason, several applications (many of them de-

scribed in the chapters of this volume) are now facing both

parties and users with questionnaires where the answer can

be placed on a 5-point Likert scale. At third, the method

has been questioned with respect to the authorization proc-

ess by political parties. StemWijzer’s methodology consists in

fact in giving to political parties the freedom to position

them on each issue without double-checking whether this

position is formally supported by the party manifesto. Ac-

cording to one of the creators of StemWijzer, authorization

by authoritative sources within the party is a sufficient

guarantee of political reliability (Schuszler et al., 2003).

Some contends that this might make the tool vulnerable for

manipulation by party elites, and campaign strategists in

particular (Groot, 2003). Therefore, some VAA makers

have opted for a process of ‘calibration’ based on a hierar-

chy of data sources, such as party manifestos or transcripts

from parliamentary debates (for a better discussion of this

point, see: Krouwel and Fiers, 2008). As a final point, draw-

ing from the literature on policy space (multi) dimensional-

ity (see: Sartori, 1976), some VAAs (e.g., Dutch Kieskompas,

Portuguese Bússola Eleitoral, Swiss smartvote) have gone be-

yond the linear notion of distance subsumed in the StemWi-

jzer methodology by placing both parties and voters on a

multidimensional space (for a better discussion, see: Lobo

et al. in this volume). These and other aspects will be thor-

oughly discussed in the following chapters.

1.6. Outline of the book

Chapter 2 by Jochum de Graaf – one of the inventors –

discusses the origins of the father of all Voting Advice Ap-

29

Chapter One - The Effects of VAAs on Users’ Voting Behaviour

plication, Dutch StemWijzer, the educational purposes un-

derlying its development and the reasons of its ‘dazzling

success’. The extensive discussion of the so-called ‘StemWi-

jzer Method’ in particular seems to us a fundamental point

of departure for the reader not particularly familiar with the

way in which VAAs operate.

Following Chapter 3 by Outi Ruusuvirta deals with the

Finnish case. As she shows, if the Netherlands are the place

in which it all began, Finland is the country where it went

online first5. Moreover, the Finnish online voting application

scene is characterized by the wider number of VAAs avail-

able to the voters, as compared to every other European

country. Ruusuvirta’s piece is of particular interest for her

discussion of the determinants of VAAs popularity among

users – which she individuates in the kind of electoral sys-

tem adopted in the country. A highly proportional system

such as that employed in Finland implies in fact the pres-

ence of many parties on the ballot. Furthermore, through

the open-list system voters are not only asked to vote for a

party, but also for one of its candidates. In this candidate-

centered context, voters are required (of course assuming

their willingness to base the vote on issue stands) to com-

pare the positions of hundred of candidates – a task easily

accomplishable through the use of VAAs. The rise of

VAAs is also linked to the spread of internet among mass

publics. Accordingly, Ruusuvirta highlights that in Finland

(as everywhere else, as we shall see) VAA usage is funda-

mentally a ‘generational phenomenon’ – the typical user be-

ing young, male and highly educated6. Finally, she reviews

5 StemWijzer’s first online version is of 1998, while the first internet-

based VAA appeared in Finland as early as 1996.

6 Unfortunately, this identikit depicts the kind of voter less likely to need

guidance in the act of voting (see Ruusuvirta in this volume for a better

discussion of this point).

30

Diego Garzia

all the available literature on the Finnish case dealing with

VAAs effects on users’ political behaviour.

The impact of VAAs on voting behaviour is at the core

of Chapter 4 by Stefan Marschall and Christian Schmidt. In

their study of an extremely popular German application

(Wahl-O-Mat) the authors distinguish between effects on

the ‘cognitive’ dimension (that is, motivating users to gather

further information about parties and candidates) and ef-

fects on the ‘activity’ dimension (such as mobilizing users

to go to the polls) of political behaviour. As they show,

Wahl-O-Mat is capable of affecting both dimensions in sig-

nificant terms. From a methodological perspective, their

piece is of extreme interest for the discussion of web-surveys

as a cheap and effective (although not widely employed)

way to study voting behaviour at the individual level.

In Chapter 5, Andreas Ladner, Gabriela Felder and Jan

Fivaz discuss the impact on voters’ choice of smartvote, a

VAA developed in the most candidate-centered system of

all Europe, Switzerland. In the Swiss case, single candidates

(rather than parties as a whole) are involved in the making-

process: the chapter’s added value lies thus in the discus-

sion of the matter also from the politicians’ point of view.

As they demonstrate, smartvote is considered ‘more than a

toy’ by both candidates and voters. Consequently, if VAAs

are to be taken seriously, they move to the crucial question:

to what extent can their providers be held accountable?

Chapter 6 by Michiel Nuytemans, Stefaan Walgrave and

Kris Deschouwer moves its steps in the same path. Being

also the Belgian a case in which VAAs (in particular, Doe De

Stemtest!) do have an impact on vote choice of a – small but

significant – number of users, the authors make a strong

point in favor of the ‘complete methodological transpar-

ency’ of VAA-making. In particular, they discuss two key

points related to the quality of the advice provided by such

31

Chapter One - The Effects of VAAs on Users’ Voting Behaviour

application: namely, the process of statement selection and

the different saliency attributed by different parties to each

of these statements7. This chapter provides also an insight-

ful conceptualization of different approaches to VAA-

making. It is the authors’ point of view that VAAs should,

first and foremost, help people vote ‘as informed as possi-

ble’. That is to say, VAAs are (or should be) aimed to help

users casting their vote, not to tell them what party they

should vote for.

The same point is shared by Marina Costa Lobo, Maarten

Vink and Marco Lisi. In Chapter 7 they present Bùssola Elei-

toral, a multidimensional VAA at work in Portugal since

2009. Contrary to every other VAA presented in this vol-

ume in fact, the Bùssola places both parties and voters on a

bi-dimensional policy space. After a thorough discussion of

the main dimensions of political competition in Portugal,

the authors describe in detail all the methodological aspects

of the application. The chapter is especially interesting for

its focus on other quality issues, such as the ‘authorization

process’. Contrary to StemWijzer in fact, the makers of Por-

tuguese VAA place political parties on issues based on an

expert judgment of party manifestos and parliamentary de-

bates’ transcripts. Rather than asking directly to parties,

their position is thus ‘calibrated’ by the makers themselves.

Furthermore, both parties’ and users’ positions are (or at

any rate, can) be arrayed on 5-point Likert scales, in order

to enhance the discriminatory power of the tool.

Chapter 8 by Christine Mayer and Martin Wassermair il-

lustrates the Austrian VAA wahlkabine.at. These authors

underline the educative purpose of such tools, which “can-

not, and should not serve as substitutes for independent re-

7 Differently from StemWijzer in fact, Doe De Stemtest! features an algo-

rithm by which the final voting advice is computed taking into account

the saliency assigned by both parties and voters to the various issues.

32

Diego Garzia

flection and responsible decision making”. Following

Chapter 9 by Roberto De Rosa presents some data from

Cabina-Elettorale.it, a pilot study conducted on Italian voters

during the campaign for the European elections of June

2009. These two chapters represent a reciprocal integration

of theoretical and pratical aspects. The Austrian and Italian

applications are in fact linked by a common methodology

and realization, both being based on the IT platform devel-

oped by the Austrian Institute for New Culture Technolo-

gies.

In Chapter 10, mathematician Michal Skop presents a

comparative assessment of quality of the advice provided

by different kind of VAAs. In his analysis of Czech Repub-

lic, he focuses on two applications: iDnes.cz and Koho-

Volit.eu. The former is based on the well-known ‘StemWi-

jzer Method’, while the latter – developed by the author

himself – relies on what he labels the ‘Roll-Call Method’.

The distinctive feature of this method lies in the statement

selection procedure: questions are based on topics which

have been voted on in parliament; parties’ position on such

topics is thus based on roll-call records from the past legis-

lature. He finds that both kind of VAA have advantages,

but also drawbacks: while ‘prospective VAAs’ such as

StemWijzer tend to rely too much on what parties propose

to do – thus ignoring what they have done so far – roll-call

based VAAs are vulnerable to the opposite critique, being

fundamentally ‘retrospective VAAs’.

Chapter 11 by Agata Dziewulska presents Glosuje.com.pl, a

VAA developed in Poland in the outset of 2009 European

Parliament elections with the aim of increasing turnout in a

country characterized by extremely low levels of electoral

participation. After having identified the major reasons of

nonparticipation (as it appears, one of the crucial argu-

ments lies in the ‘poor understanding’ of the elections by a

33

Chapter One - The Effects of VAAs on Users’ Voting Behaviour

substantial majority of Polish voters), she moves to an illus-

tration of how those reasons could be neutralized. In this

context, Glosuje.com.pl is presented as a way to provide citi-

zens with effective political information at a cheap price –

thus highlighting once more the fundamentally educative

purpose of VAAs.

Chapter 12 by Lorella Cedroni draws the conclusions and

outlines avenues for further research in the field.

CHAPTER TWO

The Irresistible Rise of Stemwijzer

JOCHUM DE GRAAF

The first StemWijzer was developed in the Netherlands in

1989. The Didactische Handleiding Maatschappijleer (in English:

Didactic Social Study Manual), published by Stichting Burger-

schapskunde (SBK: Citizenship Foundation), contained a

simple test with extensive, especially ideological, statements

from political party programs. In that year Paul Lucardie,

an employee of Documentatiecentrum Nederlanse Politieke Parti-

jen (DNPP: Documentation Centre of Dutch Political Par-

ties) came to SBK with the idea of developing a voting aid

tool – in the same sense as he had done in a limited scale in

the city council elections of Groningen. The StemWijzer –

an idea of the present author, then project leader at SBK –

was the co-production of SBK, DNPP and the faculty of

Bestuurskunde (Political Management) at the University of

Twente (UT). The StemWijzer package was a small book

with 60 statements and a diskette. The package was meant

for junior high-school education.

This first version was based on a simple principle,

namely, that all political parties can be placed on one di-

mension: a left-right scale. If one agreed with a particular

statement, this gave a range from 0 to 50 points, depending

on the party standing for the statement. By dividing the to-

tal amount of points with the amount of the statements

agreed upon, the advice was computed and given. Before

the package was advertised in the professional magazine of

social science and social and political education teachers, it

could be ordered through SBK. The book was fairly popu-

36

Jochum de Graaf

lar especially within the educational sector. However, only

around 50 copies of diskettes were sold.

Only after that, we started considering the possibility of

using computers, on the basis of the ideas of UT employee

Peter Schuszler. The first, more advanced digital StemWi-

jzer was published at the time of the Tweede Kamer Parlia-

mentary election of 1994. The reach of publicity with the

articles in nation-wide newspapers and interviews on radio

was particularly significant. Despite this tension, the use of

StemWijzer remained limited. In total, some thousand disk-

ettes were sold, partially through SBK (that by then merged

into the Instituut voor Publiek en Politiek; IPP: Institute for

Political Participation), but also via kiosks and bookstores.

In 1998, the StemWijzer for the parliamentary elections was

not only spread out in a large number via diskettes and to

the readers of the daily newspaper Trouw among others; it

was also published on the Internet on the election site of

IPP. 6.500 voters used the first Internet version.

Ever since the beginning, StemWijzer has had mostly an

educational purpose. In the teacher’s manual of 1989, this

purpose was expressed as follows: i) increasing knowledge

of the programmatic differences and similarities between

political parties; and ii) providing aid for making choice for

a political party. The aim has always been to show the po-

litical differences between parties, on the basis of the as-

sumption that voters should know these differences and be

able to compare them with their own viewpoint and politi-

cal position. This is the educational purpose of StemWijzer,

which is expressed by the name itself, which means “voting

wiser”. From the research we have conducted throughout

the years, we have found that StemWijzers are not only

played, but also used as the starting point for discussion in

family or friends, at classes, workplaces, in cafés or on the

streets.

37

Chapter Two - The Irresistible Rise of Stemwijzer

2.1. Realization

The election manifestos of the involved political parties

are at the basis of every version of StemWijzer. The first

job of the editors is to select the issues (in the form of sim-

ple statements) from different election programs. During

the selection, attention is paid to the dispersion of the

themes that are in the programs, whether positions are con-

troversial enough (at least one party against and one party

pro), and whether the positive and negative statements are

appropriately in balance. The first selection results in a list

of some hundred theses. This selection is then discussed in

a wider circle, where attention is paid to actuality, relevance,

clarity, and so on. New statements can also be added during

this time. In the end, a gross list of roughly fifty theses is

left, that are given to the recognized members of political

parties. They decide whether the party agrees, disagrees or

is neutral to the given fifty theses. Often, this time is also

used by the involved personnel to comment on the given

formulation of the theses or the lack of relevant issues.

This critique is taken into account in the final version as

much as possible. Then the final shifting is done. Theses

that do not have at least one party agreeing or disagreeing

are eliminated. Next to that, theses where quite a number

of parties have chosen for the neutral position are also

likely to be skipped. The statements should be well spread

over different themes, and all parties should remain differ-

ent enough from each other in their positions. If a combi-

nation of theses has too few difference points, then differ-

ent theses will be chosen. The goal is to finish with roughly

25 to 30 theses.

It is clear to everyone that bringing the text of hundreds

of pages of party programs down to approximately thirty

questions will always remain controversial. Every person

38

Jochum de Graaf

with a moderate interest in politics can easily think of a

number of subjects that are not dealt with in StemWijzer.

Yet, we believe that the final choice of statements repre-

sents a real summary of political issues that will play an es-

sential role in the parliamentary period ahead, and thus is

good to have knowledge of the positions of political parties

on such issues. We have a number of reasons to believe

this:

- All theses and election themes come from the election

programs of participating political parties. Every party

represents a certain number of election subjects.

- We look for theses that are relevant and characteristic of

the involved political parties, for example by looking at

what each party states in the introduction of their party

programs.

- In the selection process, the issues that are treated in the

election campaign are looked at, without easily going

with ‘the illusion of the day’. Political programs pre-

sented during election, after all, go for four years – a

parliamentary term.

Furthermore, in the selection process political parties

themselves are ultimately involved, when they answer the

gross list of theses. The positions of political parties to

these theses give party profiles to which the voters will be

compared with.

According to certain critiques, there is discrepancy be-

tween what party says to stand for, and what they actually

mean. In particular, the parties on the far side of left and

right spectrum are supposed to use this tactics. Other par-

ties might be dominantly profiling on one subject (e.g., sin-

gle-issue parties). Another tactic might be that of choosing

answers which they think people would like to hear.

39

Chapter Two - The Irresistible Rise of Stemwijzer

Every judgment of any independent political forum al-

ways carries the danger of arbitrary judgment paired with

the ever-lasting discussion about political parties’ positions.

To avoid this, we have chosen a simple solution, which is

to have the parties answer these questions themselves. The

responsibility is explicitly expressed in the colophon (Fre-

quently Asked Question) part of the website. Because of

this, StemWijzer also acts as a test of the consistency of

party programs themselves. Due to the long experience of

making StemWijzers over the years, we can in some cases

point out the inconsistencies of the standpoints of the par-

ties over time. In 1994, there were still some arbitrary an-

swers from the political parties, not being answered seri-

ously. In the following years, the authorization process of

the political parties came to be taken more seriously. It

happened several times that the political parties changed

their positions after careful consideration.

StemWijzer became an authority, and political parties

started to recognize its importance. In the past, only the

campaign leader or the drafter(s) of the party manifesto

were answering the StemWijzer questions. Nowadays, it is

not rare that the list of questions goes through the different

levels of the party. For example, in CDA, they let members

of multiple levels take tests, involving: all members of the

campaign teams, party leaders and candidate members for

European Parliament, the member of the external commis-

sion of the party – up to 40 members in total. The authori-

zation by the parties leads to constructive comments, in

some cases to extensive discussions within the parties. Be-

cause StemWijzer is placed in a public space on the Inter-

net, we also receive hundreds of emails, most of which are

positive, but also people asking why certain party takes cer-

tain standpoints. The editors of the StemWijzer forward

these questions to the respective parties.

40

Jochum de Graaf

2.2. Method of calculation

As of 1998, the StemWijzer Method is as follows: a user is

provided with the theses, to which (s)he can choose to

agree, disagree, be neutral or have no opinion1. In the case of no

opinion, the thesis is not taken into account when calculating

the voting advice. When all the theses are answered, the

user can assign extra weights to certain theses. Then the

computer compares different party profiles to that of the

user, on the basis of the principle of shortest distance (see

Table 2.1). If the user agrees with a thesis (+) and the party

disagrees (-), the difference is maximum (2 points). If a user

has the same opinion as the party (agree, disagree or neu-

tral), then there is no difference (0 point). If a user or a

party takes a neutral standpoint, the score is 1.

Table 2.1 – Scoring on the basis of ‘closest distance’

Parties:

Voter:

Disagree

Neutral

Agree

Disagree

0

1

2

Neutral

1

0

1

Agree

2

1

0

With 30 theses, the maximum distance according to

above formula is 60 points, and the smallest distance is 0.

The maximum distance can become larger if the user se-

lects one or more theses to be weighted more heavily (if the

subject is important to them). The points for these theses

are then doubled and thus get to a maximum of 4 points

1 Since 2005 the answer options are as follows: agree, disagree or no opi-

nion.

41

Chapter Two - The Irresistible Rise of Stemwijzer

per thesis. The party with the lowest score is at the top of

the recommendation list provided at the end of the test.

One of the most asked question about StemWijzer is

whether the application gives a correct or trustworthy ad-

vice. Of course, this leads to the question: what is a correct

advice? Here, it helps to have the educational aim of Stem-

Wijzer in perspective. StemWijzers are, above all, meant to

make the voters literally wiser. What StemWijzers do is no

more and no less than submitting a number of political

statements to the user, and then compare his opinions with

those of the political parties. Apart from a voting advice,

StemWijzer challenges users to think about most important

actual political questions. They get to know how parties

think about these theses and through this reach new

knowledge and insights. From recent researches, it seems

that users do pick up the idea of StemWijzer quite well.

They do not pay the most attention to the voting advice,

but are encouraged to think. Especially the ‘surprising ef-

fect’ is appreciated; recommendation that is different from

expectation is a good motive for comparison of party

standpoints and for further discussion with family and

friends. One more finding from the research is that voting

tests such as StemWijzer will play a large role in future elec-

tion campaigns, because it offers a good counter-balance

against the ‘media-democracy’ trend of recent years (see, in

this respect, the various chapters in this volume).

2.3. The reasons of a dazzling success

The StemWijzer has grown – in one-year period between

2002 to 2003 – into the most used Internet application dur-

ing election time. From 50 sold brochures in 1989 to 6.500

given advice in 1998, then to more than 2 million in succes-

sive parliamentary elections in 2002 and 2003, up to 4.7

42

Jochum de Graaf

million in 2006: dazzling success. Yet there are a number of

good explanations for this achievement. First, the ‘success’

of an election: an exciting election campaign and thrilling

political debates are necessary conditions. Something has to

be at stake for an election to receive attention. For the

voter, it has to be clear that there is something to choose

for. The election campaigns of both 2002 and 2003 did

meet all these conditions for lively democracy. The irre-

sistible rise of Pim Fortuyn and his LPF also had a huge ef-

fect on StemWijzer.

In March 2002, Fortuyn did not yet have a party program

available. StemWijzer 2002 was launched on March 10th

without the participation of LPF. 10 days later, the book De

Puinhopen van Paars (The Mess of the Parliament) which was

supposedly functioning as an election program of LPF was

published, and immediately after this LPF approved the

gross list of statements. In a very short time a new StemWi-

jzer, now featuring LPF, was brought online. This not only

provided new publicity, but also did increase the number of

visitors very much. Most part of voters wanted to know

what LPF was standing for, and what their own positions

were in respect to the controversial issues. The extremely

lively, but also turbulent campaign of 2002, increased the

turn-up rate by 5%. When the first cabinet of prime-

minister Balkenende fell after four months and a new elec-

tion was required, then again it was a very lively campaign.

Once more, StemWijzer was indispensable part of it. In

May, the number of 2 million recommendations (that is,

completed tests) was reached in 11 weeks. StemWijzer for

parliament election of 2003 came online on December 3rd

and broke the record of 2 million within 7 weeks with addi-

tion of about 200.000. As always, the number of visitors in-

creases as the election campaign progresses. On the elec-

43

Chapter Two - The Irresistible Rise of Stemwijzer

tion day of January 2002, over 200,000 recommendations

were given out.

Another major explanation is the enormous flight of the

Internet at the beginning of the century, both in the num-

ber of users, as well as its general usage. In 1998, roughly 1

million people in the Netherlands had access to the Inter-

net. In years 2002 and 2003, this number was roughly ten-

fold of that.

StemWijzer is not only attractive as an Internet website,

but also as an independent voting advice tool. This was the

case when StemWijzer took part in the TV program Studio

2 during the campaign of year 2002, where guests received

their voting advice and from there went into debates. In the

campaign of 2003, similar arrangements were made with

the radio program Stand.nl and with Freek de Jong for his

broadcasted TV theatre-show De Stemming (The Vot-

ing/The Mood). In the 2006 campaign there was, exactly

one week before Election Day, a 90 munites tv-show

StemWijzer TV, with debates on the theses by party leaders

and the audience at home participating by answering the

statements by SMS.

The success of StemWijzer also comes in large part from

its design, the presentation, the user-friendliness and

handiness of the program. Most voting tests are character-

ized by a multiple-choice set-up, simple layout, long ques-

tion lists on scroll-down Internet site, clicking on radio but-

tons to give the answers. With StemWijzer, the theses are

presented one by one in a tight layout, where the answering

categories are presented with clickable buttons. In this way,

the visitors can digest the theses well and take considered

positions.

44

Jochum de Graaf

2.4. The future of StemWijzer

From the beginning of StemWijzer, there have been dis-

cussions about the set-up of the program. Each time, the

reflected point is if the answering categories are sufficient,

the question of whether the five-scale-answer of between

fully agree and fully disagree would have to be introduced,

or whether for example a pop-up window with explanation

of difficult terms should be given, if the political parties

should be given the opportunity to comment further on

their standpoints with a link to their party program, or if

questions with multiple-answer could also be asked. These

suggestions of improvement should never be shut out from

the possibility of implementation; however, we chose so far

to stick to the simplest possible set-up. The propositions

are presented one by one, the vote advice can be compared

with on the level of a party and on the level of proposi-

tions. From all our researches, it appears that the users ap-

preciate this compact set-up, next to the curiosity for the

voting advice and the surprise-effect of StemWijzer, which

are all are highly valued.

In recent years the StemWijzer has improved itself to an

even more elaborated educational tool. Throughout the

years the calculation method has been profoundly dis-

cussed. The method mentioned in Table 2.1, on the ‘closest

distance’ (or ‘city-block method’) has been used in different

variations. Especially the valuation of the neutral or ‘mid-

dle’ position, was a few times changed; the so-called

Cohen’s kappa coefficient was used, the statistical measure

of inter-coder agreement for qualitative (categorical) items.

In principle this is a more robust measure since it takes into

account the agreement occurring by chance. But this work-

ing quite well for scientific research did not mean that it

should work for a simple educational test like StemWijzer.

45

Chapter Two - The Irresistible Rise of Stemwijzer

When used in the campaign for provincial elections in

2007, this method proved to be disadvantageous for parties

(since we were using the method for a selection of party

answers that was chosen by us), thus we decided to get

back to the most simple calculation method. Nowadays the

result (e.g., voting advice) of the StemWijzer is calculated

by simply counting the agreements between the answers of

the users and those of the parties. Each answer that has

been given an extra weight does get an extra point.

An even more important improvement for StemWijzer is

the addition of the motivations of parties to the answers

given. As from the parliamentary elections in 2006, when

the all-time users record was obtained, the user could click

on the party answer and consult in a pop-up screen the ar-

guments from the party to take side pro or contra. In all re-

cent versions of StemWijzer (European elections 2009,

municipal elections of 2010, where some sixty different

StemWijzers for all major cities in the Netherlands have

been developed) this functionality is available on every the-

sis. When thinking about an answer, the user can, by click-

ing on a party logo, check all deliberations of the parties. By

strengthening the educational value of the tool, the user can

vote even wiser.

In the upcoming election campaigns in the Netherlands,

we will also continue the experiments in having users to

participate in the selection of issues and the formulation of

theses. In some municipalities StemWijzer fans could

change and add theses with the Wikipedia-method. In the

near future a forum application will be added, as well as

discussion groups from social networks like Facebook or

Hyves. In this way StemWijzer will be ready for the next

developments in the lively world of internet and politics.

46

Jochum de Graaf

In the end of year 2003, StemWijzer was awarded the Machiavelli

prize, the yearly distinction award for the performance in the field of

government communication. The jury report speaks of a ‘beacon in the

sea’, the ‘ocean stream’ of information in election times: “In a time of

the ongoing mediatizing, dramatization and personalization of politics

in general and election issues, StemWijzer offers a program-content-

based politics to counter-balance the above trends”.

CHAPTER THREE

Much Ado About Nothing?

Online Voting Advice Applications in Finland

OUTI RUUSUVIRTA

The first online voting advice application (VAA) in

Europe was developed in Finland in 1996 for the country’s

first European Parliament Election. Erkki Vihtonen, a pro-

ject manager at the Finnish Public Broadcasting Company

Yleisradio (YLE), had been inspired by a simple online

game on the website of the US news corporation CNN and

decided to develop a tool to inform Finnish voters about

the European Parliament Election candidates and help

them choose a suitable candidate (Vihtonen, 2007). From

this humble beginning with only 8.000 users has grown a

phenomenon that is now a permanent fixture in the Fin-

nish electoral landscape.

This chapter will review the short history of VAAs in

Finland. We begin by tracing the development of Finnish

VAAs before turning to the examination of the factors that

make these applications so popular in the country. We will

then analyse profiles of Finnish VAA users. Finally, we will

conclude with a review of research conducted on the ef-

fects of online voting advice applications on their users.

3.1. Development of Online Voting Advice Applications in Finland

Many features distinguish the Finnish online voting appli-

cation scene from the rest of Europe. One such feature is

48

Outi Ruusuvirta

the large number of VAAs available to voters. As noted

above, the Finnish Public Broadcasting Company devel-

oped the first VAA in 1996 and the largest daily newspaper,

Helsingin Sanomat (HS) followed this example by building

its own application for the 1999 European Parliament elec-

tion. The first Helsingin Sanomat VAA attracted 15.000 us-

ers (Haukio and Suojanen, 2004: 129). YLE and HS appli-

cations were still the only two VAAs available in the 2000

Presidential election, where they were used by 150.000 and

65.000 citizens respectively (ibid.). The following year, 11

different VAAs were developed for the local election (Suo-

janen, 2007: 17). The first two pioneering applications were

joined by VAAs from the biggest commercial TV channel,

MTV3, the Association of Finnish Local and Regional Au-

thorities and a large number of regional newspapers.

The real breakthrough for voting advice applications

came in the 2003 Parliamentary Election. This time 12 dif-

ferent VAAs offered their help to voters. Previously, the

audiences reached by the most popular VAAs had ranged

between 100.000 and 200.000 users, depending on an elec-

tion, but in the 2003 General Election the two largest

VAAs reached approximately 530.000 and 410.000 users

respectively (Kauppinen, 2007: 127; YLE Viestintä, 2003).

Considering that the size of the electorate in these elections

is approximately 4.2 million, it can be argued that VAAs

truly established themselves as a part of election campaigns

that year. Since 2003, voting advice applications continued

to grow both in number and in popularity. The 2004 EU

Election and the 2006 Presidential Election saw many new

VAA providers offering their VAAs to the electorate,

among them political parties building their own applica-

tions to market their candidates to voters and interest or-

ganisations offering applications concentrating on specific

themes.

49

Chapter Three - Much Ado About Nothing? Online VAAs in Finland

In the 2007 Parliamentary Election voters could choose

from 20 different VAAs. The most popular applications at-

tracted over one million users (Borg, 2008: 95). Several

VAAs targeting special audiences, such as young voters,

those interested in development issues or health care, also

appeared in this election. Up until this point, Finnish VAAs

had more or less followed the basic format first used in

1996 by presenting users with a multiple-choice question-

naire to compare their preferences with candidates’ views.

In 2007, a group of young mathematicians developed a

completely new type of VAA. This Naama application

compared user’s picture to those of the candidates’ and

matched them based on the degree of facial similarity1. The

number of applications went through the roof in the 2008

local election as many local and regional newspapers and

local authorities provided VAAs concentrating on local

themes and issues to supplement the more general national

applications. Today, online applications helping voters to

choose have also spread beyond national elections and are

used in a wide variety of contests from university student

council elections to parish council elections.

Simply looking at VAA user figures does not tell us the

whole story about the popularity of these tools among the

electorate at large. As Mykkänen and his colleagues point

out in connection with the 2007 General Election, 63 per

cent of those who used VAAs took advantage of more than

one application (Mykkänen et al., 2007). Other studies have

suggested that this has also been the case in other elections

(Bengtsson and Grönlund, 2005: 247; Paloheimo, 2007:

1 In the 2007 elections 167.000 users compared their images to candi-

dates. The creators of Naama have continued to develop facial similari-

ty VAAs, not only for Finnish elections but also for the 2008 US Presi-

dential elections. The applications are available at

<http://www.naama.fi/index.php?lang=en>.

50

Outi Ruusuvirta

61). We must therefore consider national election study

data for a more accurate picture. Voting advice applications

were first included as an independent variable in the 2003

Finnish Election Study. Before that, the study asked re-

spondents only about their Internet use in general. Of all

respondents in the 2003 study, only 12 per cent had used at

least one VAA in the run up to the election. When we con-

sider respondents who actually voted, the figure grows to

16 per cent. Of those who both voted and used the Inter-

net during the campaign, 40 per cent had used at least one

VAA (Paloheimo, 2007: 59). On the whole, these figures

appear rather low. However, as will be shown below, VAA

use is conditioned by other factors, especially the age of the

voter, and in certain cohorts use of these tools is much

more common than in the electorate as a whole. For in-

stance, of the 18-24 year olds, 46 per cent reported to have

used at least one VAA during the 2003 campaign (ibid., 60).

As with the number of VAAs themselves, the share of the

electorate using them also grew in the 2007 General Elec-

tion. Of all respondents, 41 per cent had used voting advice

applications and in the two youngest cohorts, nearly 59 per

cent used VAAs (Mykkänen, 2009, personal communica-

tion). It would then appear that by the 2007 Election,

online voting advice applications had reached a large

enough proportion of the electorate to become a significant

factor in electoral analysis.

While such a large number of applications is arguably

good for the democratic process as voters have several

VAAs to choose from, Finnish election candidates are be-

ginning to complain about VAA fatigue. While very few

would dare not to respond to the three most widely used

applications along with perhaps one regional and a special

interest VAA, many are expressing their frustration about

the time and effort taken in filling out numerous voting ad-

51

Chapter Three - Much Ado About Nothing? Online VAAs in Finland

vice applications during the already busy campaign period

(Vähämaa, 2007). Nevertheless, it appears that online vot-

ing applications have become a permanent feature of Fin-

nish election campaigns. Applications rate as the most

popular online information source for voters (Strandberg,

2009: 77) and VAAs form the most popular election-

specific content on the websites of media houses providing

them. As one VAA provider put it: “It would be hard to

imagine an election campaign without VAAs”2. Despite

questions raised in blogs, other online forums and coffee

tables about validity and reliability of VAAs, their unknown

matching algorithms, question selection and response alter-

natives, the open-list proportional representation (PR) sys-

tem used in Finnish elections along with other factors make

VAAs an easy way for voters to find comparable informa-

tion on election candidates (Borg, 2008: 95). Let us now

consider these factors more closely.

3.2. Why Are Online Voting Advice Applications Popular in

Finland?

Finnish parliamentary, European Parliament, and local

elections all employ the open-list PR system in multi-

member districts3. This means that voters do not only vote

2 O. Ainola - Interviewed on 23 February 2009, Helsinki.

3 Finnish President is chosen directly by popular vote using a two-

round majority system. Typically, the number of candidates is low

(eight in 2006) and the number of top candidates with a realistic chance

of being elected is even lower. Leading presidential candidates are also

well-known political figures and therefore VAAs in these elections have

concentrated somewhat less on political issues and more on the perso-

nality and non-political views of the candidates. It is possibly for these

reasons that VAAs are used less in Presidential elections than in par-

liamentary and local elections. For instance in the 2006 election, only a

52

Outi Ruusuvirta

for a party but also have to select a candidate from a party’s

list to vote for. Candidates are then ranked on the party

ticket according to the individual votes they receive. This

leads to a candidate-centred election dynamic. While such

an electoral system gives a large degree of choice to voters,

it also makes the selection process harder for them; instead

of choosing from a handful of parties, voters have to

choose from hundreds of candidates. Even accepting that

no voter will give equal consideration to all the candidates,

a choice can still be bewildering and information costs very

high (cf. Downs, 1957). Voters therefore use shortcuts to

help them decide how to vote. Traditionally, the electorate

relied on long-term cues such as party identification

(Campbell et al., 1960) and social class for electoral cues

(Lipset and Rokkan, 1967). As is well-known, the impor-

tance of these factors for electoral choice has declined in

most advanced post-industrial democracies in the past 40

years (Franklin et al., 1992). According to the 2003 Finnish

Election Study, the share of those feeling close to a political

party had declined to 47 per cent of the electorate

(Grönlund et al., 2005: 100).

However, in the Finnish electoral system partisan or class

alignment could only ever have been expected to narrow

down the pool of potential candidates a voter would con-

sider. Even the strongest of partisans must select an indi-

vidual candidate to vote for. When asked for the criteria

used to select a specific candidate, a large majority (82 per

cent) refer to candidate’s views on general political and

election issues4 (Bengtsson and Grönlund, 2005: 244-5).

quarter of voters had used at least one VAA during the election cam-

paign (Nieminen, 2006).

4 Other selection criteria included the party represented by the candi-

date (74 per cent), candidate’s previous experience of politics (56 per

cent), his or her educational background (37 per cent), gender (33 per

53

Chapter Three - Much Ado About Nothing? Online VAAs in Finland

This figure may overestimate the real influence of issue

considerations, given social desirability voters may feel is

attached to giving such an answer. Nevertheless, it does

suggest that issues play a part in voters’ candidate selection

in Finland and amassing detailed information on potential

candidates even on a few key issues would require enor-

mous amount of effort from even the most politically so-

phisticated voters. Thus, the candidate-centred election dy-

namic combined with voter dealignment amounts to a

situation where more and more voters are genuinely mak-

ing a choice at election time and online voting advice appli-

cations, which easily allow comparison of individuals’ pref-

erences with those of the candidates, can be very useful for

voters. Indeed, in a nationally representative survey of both

VAA users and non-users, Mykkänen et al. found 73 per

cent of VAA users agreeing that the applications were a

good way to learn about candidates’ views and 28 per cent

agreed that VAAs made selecting a suitable candidate eas-

ier. On the other hand, 20 per cent of VAA users had used

a voting advice application to help rule out a candidate they

had considered voting for (Mykkänen et al., 2007).

Related to the electoral system, another feature of the

Finnish political landscape that can be used to explain the

emergence and the success of online voting advice applica-

tions is the relatively high number of electorally relevant

political parties. The Finnish party system is the most frag-

mented in Western Europe. The mean number of effective

parties between 1945 and 1999 was 5.1 (Mattila and

Raunio, 2004: 269). The three largest parties, the National

Coalition Party, the Centre Party and the Social Democratic

Party, each command approximately 20 per cent of the

cent), appearance and style (32 per cent), how well-known the candi-

date was (32 per cent) and his or her age (28 per cent) (Bengtsson and

Grönlund, 2005: 245).

54

Outi Ruusuvirta

vote in parliamentary elections. In additions to the big

three, five smaller parties also regularly win seats in the

Finnish Parliament, Eduskunta. On the left of the political

spectrum, the Left Alliance and the Green League chal-

lenge the Social Democrats, whereas on the centre-right,

the Christian Democratic Party, the Swedish People’s Party

and the populist True Finns compete for voters’ affections

with the National Coalition Party and the Centre Party.

Thus, Finnish voters on both sides of the political spec-

trum have several parties to choose from. The large num-

ber of parties combined with the candidate-centred elec-

toral system described above leads to a situation where for

instance, voters in the Uusimaa electoral district could

choose from 340 candidates competing in the 2007 General

Election. It is no surprise then that online voting advice

applications have become so popular in Finland.

As we noted in the first section of this chapter, media

companies have developed a majority of Finnish online

voting advice applications. The way in which these news-

papers and TV channels have promoted their VAAs both

in the online world but especially in their traditional media

format should also be taken into account when explaining

the popularity of VAAs in Finland. Naturally, voting advice

applications have been given plenty of visibility on news

providers’ websites and links to VAAs are widely available.

Newspapers also print advertisements in their paper ver-

sions advertising the applications. What the Finnish media-

based VAA providers have become especially skilled at is

their use of data from voting advice applications. Journal-

ists take advantage of the candidate data deposited in VAAs

and use it as a source of news stories for their traditional

media format (Pitkänen, 2009: 122). Both newspapers and

TV news regularly run stories comparing candidates’ views

and preferences on key election issues and investigating

55

Chapter Three - Much Ado About Nothing? Online VAAs in Finland

party cohesion by aggregating candidate responses. These

stories, it could be argued, attract even more voters to use

VAAs and appeal perhaps especially to those voters who

would not otherwise use these applications.

Finally, we could hardly explain the popularity of online

voting advice applications in Finland without reference to

the widespread use of the Internet. Being a web-based tool,

VAAs cannot be used without access to the Internet. In the

spring of 2009, 78 per cent of Finnish households had

Internet access and 82 per cent of 16-74 year old Finns re-

ported to have used the Internet in the last three months. A

large majority of Internet user, 82 per cent, go online every

day or almost every day. As could be expected, the usage is

slightly skewed towards younger people; nearly 100 per

cent of under-34 year olds use the Internet. We see a small

drop in Internet usage in the age group 55-64 (just under 70

per cent) and a bigger drop in the oldest cohort of 65-74

year olds (33 per cent) (Tilastokeskus, 2009).

3.3. Who Uses Online Voting Advice Applications in Finland?

Strandberg (2009) analysed the 2007 Election Study data

to find out which factors best explain Finnish voters’ VAA

use. Predictably, the young, the politically active and those

interested in politics are most likely to use online voting

advice applications (Strandberg, 2009: 80).

In the light of what is known about the relationship be-

tween age and the strength of political preferences (e.g.

Converse, 1969) and the Internet user figures presented

above, it is hardly surprising that the most eager VAA users

are the young. The Finnish National Election Study shows

that in the 2003 Parliamentary Election nearly a half of vot-

ers below the age of 35 had used at least one VAA during

the election campaign. The share of VAA users among

56

Outi Ruusuvirta

those who had voted declines the older the voters get. In

the age group 35 to 44, 34 per cent had used at least one

VAA, whereas 15 per cent of 45 to 54 year olds and 10 per

cent of 55 to 64 year olds had done the same. Only five per

cent of over 65 year olds had used VAAs in that election

(Bengtsson and Grönlund, 2005: 246-7). The questions

used in the 2007 Election Study are not fully comparable to

those used in the 2003 Study but we can nevertheless see

that VAA use had increased in all age groups. While those

under the age of 35 are still the most frequent VAA users,

the older cohorts have caught up with the younger ones.

Only 24 per cent of under 35 year old voters say that they

have never used VAAs. In the 35 to 49 age group, 47.6

percent report never to have used a VAA whereas 53 per

cent of voters over the age of 50 had not done this either5

(Mykkänen, 2009, personal communication). It would seem

that VAA use is a generational rather than age-related phe-

nomenon. The younger cohorts with low partisan identifi-

cation use VAAs when they first become eligible to vote

and continue to use these applications as they get older

(Moring and Lindfors, 2005).

Although data from the 2003 Election Study and research

conducted after the 2004 EU election suggest that socio-

economic factors are significant in predicting VAA use

(Bengtsson and Grönlund, 2005: 247-50; Moring and Lind-

fors, 2005: 7), voter’s level of education and household in-

come were not found to be significant explanatory factors

for VAA use in the 2007 study (Strandberg, 2009: 78-81).

The difference may be explained by the rapid growth in the

5 A similar age-related pattern in VAA use is also found in local and

European Parliament elections. For instance in the 2004 European

elections, 57 per cent of 25 to 34 year olds used at least one VAA and

the 18-24 cohort was the second most avid VAA user group (Moring

and Lindfors, 2005).

57

Chapter Three - Much Ado About Nothing? Online VAAs in Finland

share of population with access to the Internet. Internet ac-

cess is now common with households from all socio-

economic groups and even those who cannot go online at

home can and do access the Net at work, school and public

places such as libraries. These findings suggest that there is

no significant digital divide between the wealthy and edu-

cated on one hand and the poor and uneducated on the

other as far as VAA use is concerned. Instead, a divide

emerges between those already interested and active in poli-

tics and those who do not find politics interesting and en-

gaging (Strandberg, 2009). This raises important questions

about the potential of online voting advice applications to

increase turnout and to help those voters with the least po-

litically relevant information to make more informed elec-

toral choices.

3.4. Effects of Online Voting Advice Application Use on Turnout and

Vote Choice

Establishing a causal relationship between VAA use and

voter mobilisation in one hand and VAA use and vote

choice on the other would require an experimental design.

However, only one such study has been conducted thus far

(Ruusuvirta, 2010). The existing research on VAA effects

relies on voter survey data and thus voters’ subjective self-

evaluations. Such evaluations can be flawed for many rea-

sons. For instance, it may be difficult for a voter to evaluate

the extent to which each possible factor affected her deci-

sion to vote and her vote choice. VAA use can also have a

subconscious or an indirect effect on voters and the share

of those who explicitly identify VAAs as the deciding factor

is likely to be rather small. Similarly, for reasons of social

desirability, a voter may feel embarrassed to admit that an

58

Outi Ruusuvirta

online tool has affected her vote choice and thus deny its

effect when asked in a survey.

With these limitations in mind, let us now consider the

effects of VAA use on citizens’ electoral mobilisation and

vote choices. We could expect VAAs to mobilise citizens to

vote by providing them with succinctly presented, personal-

ised information on candidates and parties running in any

given election (cf. Delli Carpini and Keeter, 1996). On the

other hand VAAs could be expected to affect voters’ elec-

toral choices by offering a choice for undecided voters,

strengthening existing preferences of those who have al-

ready decided by confirming the ‘correctness’ of their

choice or by convincing decided voters to change their

preference (for a longer theoretical discussion on VAA ef-

fects, see: Ruusuvirta and Rosema, 2009).

Mykkänen and Moring (2007) analysed VAA effects on

turnout in the 2003 Parliamentary and the 2004 European

Parliament elections. Using probability analysis, they find

that VAAs can significantly boost the likelihood of voting

among those of lower socio-economic status (SES) but

make hardly any difference for voters from higher socio-

economic groups. The probability to vote in the 2003 Par-

liamentary Election for men with low SES increases by 21

percentage points from 56 per cent to 77 per cent after us-

ing an online voting advice application. The change for

women from lower socio-economic backgrounds is equally

impressive. The probability to vote increases by 23 percent-

age points, from 53 per cent to 76 per cent. The mobilising

effect for both men and women from higher socio-

economic backgrounds is only 1 percentage point, from

98% to 99% (Mykkänen and Moring, 2007: 23). VAAs also

had a mobilising effect on both men and women from

lower socio-economic classes in the 2004 EU election. The

probability to vote increased by 17 percentage points for

59

Chapter Three - Much Ado About Nothing? Online VAAs in Finland

men (from 34 per cent to 51 per cent) and by 16 points for

women (from 29 per cent to 45 per cent). As with the 2003

General Election, VAAs did not significantly boost the

probability that men or women of higher socio-economic

status would vote in the 2004 European election. Both

sexes received only a 2 percentage point boost, with their

likelihood of voting increasing from 96 per cent to 98 per

cent (ibid.). Mykkänen and Moring do not offer an explana-

tion for the lack of a booster effect among the higher

socio-economic classes but “it can be speculated that voters

with higher SES were more interested in the election, knew

more about the parties, candidates and election issues and

were better able to use this information to make their

choice than their less well-off counterparts even before us-

ing advice sites. Consequently vote selector use does not

provide additional mobilising incentives for them” (Ruusu-

virta and Rosema, 2009: 8). We could also have expected to

find a bigger mobilisation boost in the European Parlia-

ment election than in the national parliamentary election as

EU elections have a lower profile and therefore voters may

have more to learn from VAAs in those elections. Myk-

känen and Moring (2007) explain this with reference to

high over-reporting of voting in EU elections. They also

argue that “voters in [Finnish] EU elections are predomi-

nantly sympathetic to the union… and more likely to vote

anyway”, even without VAAs (ibid., 20).

A less direct and more uncertain measure of the mobilis-

ing effect of VAAs is to ask citizens whether online voting

advice applications make them interested in politics and

elections. The obvious problem is that we do not know

whether the increased interest translates to voting. Never-

theless, a poll conducted after the 2007 General Election

found that 36 per cent of VAA users agreed that these web

applications made them interested in politics and elections

60

Outi Ruusuvirta

(Mykkänen et al., 2007). This figure may sound rather low

but, as discussed above, we must bear in mind that the ma-

jority of VAA users are already interested in politics and

elections and therefore it would be more surprising if the

figure was indeed higher.

When we consider VAA effects on vote choice, we find

that of those who had used at least one VAA during the

2003 General Election campaign, 30 per cent reported that

VAA recommendations had affected their candidate choice

a lot or quite a lot. This means that only seven per cent of

all voters admit that VAAs had had an impact on their can-

didate choice (Bengtsson and Grönlund, 2005: 246-50). A

data available for the 2007 Election is not fully comparable

with the 2003 Election Study but the results point to the

same direction. Thirty-four per cent of voting advice appli-

cation users agreed that VAAs had affected their candidate

choice. This translates to 19 per cent of the entire elector-

ate agreeing that VAAs impacted on their candidate choice.

A surprisingly high 15 per cent of VAA users even admit-

ted that they had no favourite candidate and simply voted

for a candidate suggested by a VAA (Mykkänen et al.,

2007). However, as we have noted above, voters’ subjective

estimates of VAA effects are likely to underestimate the

real impact these applications have on their users’ electoral

choices. The effect is also a matter of interpretation. Fin-

nish VAAs typically list all the suitable candidates in the or-

der from the most suitable to the least suitable. Some might

only say that a VAA affected their vote choice if they end

up voting for the candidate whom the application sug-

gested as the most suitable whereas others might say the

same if the candidate they vote for appears somewhere at

the top of the recommendation list.

Another way of evaluating the effects of VAA use on

vote choice is to look at the share of voters who identify

61

Chapter Three - Much Ado About Nothing? Online VAAs in Finland

VAAs as an important information source for their elec-

toral decision-making. In the 2003 Election Study, only 17

per cent of all respondents said that they had found a lot or

quite a lot of electorally significant information in VAAs.

As such, VAAs rank below most other traditional informa-

tion sources, such as newspapers, news and current affairs

programmes on TV, election advertising and family mem-

bers. However, when we consider only those respondents

who had used VAAs, the figure increases to 65 per cent

(Paloheimo, 2007: 65-6 & 71-2). This suggests that those

who actually use VAAs do find them an important source

of information. The significance of VAAs as an informa-

tion source does not appear to have changed by the 2007

General Election. A Gallup Kanava Study conducted by

TSN Gallup found that across the whole electorate, 17 per

cent of respondents said that VAAs were very or quite im-

portant for their electoral decisions (Strandberg, 2009: 82).

However, in the 2007 Election, those who had used online

voting advice applications rank them as the most important

source of electorally relevant information, even above the

traditionally important newspaper and TV coverage of cur-

rent affairs6 (Mykkänen et al., 2007). As we would expect,

the youngest cohort finds VAAs most important for their

decision-making. Fifty-five percent of 18-24 year olds say

that VAAs were very or quite important information source

for their vote choice. A third of 25 to 34 year olds and

nearly a quarter of 35 to 44 year olds also feel that they had

received a lot or quite a lot support for their electoral deci-

sion from VAAs. Above the age of 45 the importance of

6 This was also the case in the 2004 local elections; voters in all age

groups identified VAAs as the most important information source

(Vähämaa, 2008). This, however, is not surprising, given the second-

order nature of local elections and the low-information setting in which

they take place.

62

Outi Ruusuvirta

VAAs in supporting electoral decision-making declines to

single digit figures (Strandberg, 2009: 82).

Above we have discussed what could be termed direct

VAA effects on citizens’ electoral behaviour. In other

words, we have considered how using online voting advice

applications affect the mobilisation and vote choice of the

person using the VAA. However, several studies both in

Finland and elsewhere in Europe have found that VAA use

makes those using them discuss the results with their

friends and acquaintances. For instance, Mykkänen and his

colleagues find that 42 per cent of VAA users had dis-

cussed their VAA recommendations with others (Myk-

känen et al., 2007). Marschall (2005) finds similar results in

Germany. Given, for instance Huckfeldt and Sprague’s

(1994) findings about the power of social communication,

it is not far-fetched to speculate that these discussions may

prompt some non-voters to use VAAs which in turn may

bring them to the polls. Current data does not allow us to

test this hypothesis but future VAA studies could include

components to study this intriguing proposition.

3.5. Future of Online Voting Advice Applications in Finland

Political scientists have been divided in their views on the

potential of the Internet to mobilise and inform electorates.

Even the first Finnish book on online voting advice appli-

cations (Suojanen and Talponen, 2007) had as its title a

word play, which could be interpreted either as ‘The Pow-

erless VAA’ or ‘The Unruly VAA’. Nevertheless, in this

chapter we have seen that, at least under certain circum-

stances, online voting advice applications are not just fun

games without any political consequence but can encourage

certain types of citizens to vote. Approximately a third of

VAA users also say that using these applications has af-

63

Chapter Three - Much Ado About Nothing? Online VAAs in Finland

fected their vote choice. We might expect a share of those

whose electoral decisions are affected by online voting ad-

vice applications to increase in the future. The young peo-

ple who over the past decade have been socialised into

electoral politics by VAAs can be expected to continue us-

ing these applications and their relative proportion in the

electorate increases year by year. In other words, the VAA

generation will gradually replace the oldest cohorts who are

the most partisan and the least likely to use online tools.

Online voting advice applications may or may not come

to replace newspapers and the television as the main chan-

nel through which the electorate becomes informed about

politics. In any case, we should bear in mind that the in-

formation citizens receive from these different sources is

different. Unlike newspapers, television, radio and advertis-

ing, VAAs provide voters with specific, personalised and

easily comparable information about issues and candidates’

and parties’ views on them. This information can poten-

tially have different effects on those receiving it than the

traditional mass media coverage and advertising. However,

the old challenge of bringing politics to those who are not

interested in it still remains. As this chapter has shown yet

again, those who would benefit the most from the informa-

tion in online voting advice applications are the least likely

to seek it.

CHAPTER FOUR

The Impact of Voting Indicators: The Case of the

German Wahl-O-Mat1

STEFAN MARSCHALL, CHRISTIAN K. SCHMIDT

More than seven years ago, several weeks before the elec-

tions to the 15th Deutscher Bundestag (the German parlia-

ment), the “Wahl-O-Mat” was launched for the first time.

Back then, in 2002 only a few people expected that this

would be the birth of one of the most successful tools in

the field of online communication in Germany. Since its

first use the tool has been deployed at every federal and

European election as well as at many elections on the sub-

national level.

The Wahl-O-Mat has become one of the most popular

online tools in the field of civic education. The European

Wahl-O-Mat which went online a few weeks before the

elections to the European Parliament in June 2009 was

used about 1.5 million times. The latest national version of

the Wahl-O-Mat deployed at the Bundestag election in Sep-

tember 2009 was played even 6.7 million times.

The Wahl-O-Mat is a non-party service, hosted by an in-

stitution that by definition is guided by a ‘supra-party’ mis-

sion: the Bundeszentrale für politische Bildung (Federal Agency

for Civic Education). The BPB is a unique institution, not

found in other European countries.

1 This chapter is a revised and updated version of a paper published

2008 in David Farrell and Rüdiger Schmitt-Beck (eds.), Non-Party Actors

in Electoral Politics. The Role of Interest Groups and Independent Citizens in

Contemporary Election Campaigns, Baden-Baden: Nomos.

66

Stefan Marschall - Christian K. Schmidt

Attached to the Ministry of the Interior, the BPB is a

governmental organization, but it operates differently than

other government agencies. Its work is supervised by a

group of parliamentarians that reflects the plurality of the

German Bundestag. Its mission is to foster Germany’s de-

mocratic culture, to inform about the topics and proce-

dures of the political process and to mobilize political par-

ticipation2.

Given the BPB’s mission, the Wahl-O-Mat as one of the

most prominent tools produced by the Federal Agency is

confronted with a range of questions: What kind of effects

does the Wahl-O-Mat have on those using it? Does it have

an impact on the extent and quality of political participation

in Germany?

Answers to these questions will provide insights into

whether voting indicators have the capacity to mobilize

people for example to go to the election booths. If tools

such as the Wahl-O-Mat could indeed contribute to voters’

mobilization, this would make a strong case for promoting

the establishment of new voting indicators, as a decline in

the election turnout has to be observed in many systems

and on different levels.

Take, for example, the German case: an overall decline of

voter turnout in Germany over the course of the last

twenty years has stimulated an intensive discussion about

effective ways to motivate citizens to take advantage of

their right to vote3. Whereas electoral turnout at the federal

level still tends to be relatively high (at least compared with

other European countries or the United States), the corre-

2 For details compare the ordinance of the institution, in:

<http://www.bpb.de/die_bpb/The_Federal_Agency_for_Civic_Educ

ation.html>; accessed 10/4/2006.

3 See: Falter and Schoen, 2005; Rattinger et al., 2006; Ohr et al., 2009;

Lewis-Beck, 2008.

67

Chapter Four - The Impact of Voting Indicators: The Case of Wahl-O-Mat

sponding proportions at the sub-national level (Länder) ap-

pear to be in continuous decline (Kersting, 2004; Fritz

2007). This gives rise to concerns about a weakening of the

democratic culture in the Federal Republic of Germany.

Can the Wahl-O-Mat help reversing this trend?

4.1. How does the Wahl-O-Mat function?

Given that the impacts of the Wahl-O-Mat can hardly be

understood without knowing about how it operates, we

must first outline its principal features. The Wahl-O-Mat is

the German adaptation of the Stemwijzer. It basically works

like its Dutch counterpart, although it has some new and

unique features. Similar tools have been implemented in

other countries, like Austria, Belgium, Great Britain, Italy,

and Switzerland (Walgrave et al., 2008a; Ruusuvirta and

Rosema, 2009). However, in quite a few respects the Wahl-

O-Mat is different from its sister applications in other

countries.

The tool confronts the Internet users with 30 to 40

propositions, such as “The German military should leave

Afghanistan immediately” or “A nationwide minimum

wage should be introduced”. The propositions, or so-called

“theses”, are the product of a group of first or second time

voters, all under the age of 27 – the so-called “Wahl-O-Mat

editorial staff”. For each Wahl-O-Mat a new editorial staff

is established. Based on the party platforms provided at the

time of the first staff meeting, the participants attempt to

locate issues in the election campaign that might be contro-

versial among the parties. Journalists and political scientists

advise them during this task.

The propositions the editorial staff looks for must meet

several criteria: above all, they should address relevant

questions in citizens’ lives; additionally, the parties should

68

Stefan Marschall - Christian K. Schmidt

have different points of view in respect of the topics cho-

sen; and they should be easy to understand. Reducing the

complexity of political discussions and policy processes

into a single easily understandable proposition to which us-

ers and parties can only express complete agreement or dis-

agreement is one of the most substantial steps for making

this tool work.

At a certain point of the preparatory process, parties are

asked to position themselves around approximately 80 the-

ses chosen by the Wahl-O-Mat editorial staff and meeting

the criteria mentioned above. The staff selects 30 to 40

propositions based on the parties’ responses. Statistical cal-

culations guide the final choice by ensuring that the se-

lected theses are able to distinguish appropriately between

the parties taking part in the Wahl-O-Mat.

As mentioned above, when the tool is finally deployed,

the parties have already positioned themselves with respect

to the 30 to 40 propositions. All party lists which have been

admitted to the elections are invited to take part in the tool.

At the most recent federal election, 29 of the 32 parties

which were admitted answered to the theses.

Responding to the theses is exactly what the Internet us-

ers can do once the Wahl-O-Mat is online: by clicking one

of three buttons (“I agree”, “I disagree” or “neutral”) they

take a stand on the propositions, or alternatively can skip

those they have no position on. Having voted on all items

of the list, users can mark propositions they consider per-

sonally important to them, giving them a special weight in

the final calculation.

At the end of the session, the Wahl-O-Mat processes the

results. It displays the distance between the single user and

the parties in the tool by showing the best fit (e.g., the party

closest to the user’s position). Additionally, it calculates the

extent of agreement between the user and all remaining

69

Chapter Four - The Impact of Voting Indicators: The Case of Wahl-O-Mat

parties by displaying the summed distances. The Wahl-O-

Mat computes the gap between the user’s and the party’s

position using the “city block” method4.

Furthermore, Wahl-O-Mat users can take a closer look at

the relationship between the positions of the parties and

their own points of view displayed for each proposition.

And the onliners have the possibility to look at the explana-

tions the parties provide to explain their positions in re-

spect to the different topics.

Working this way, the Wahl-O-Mat has become an out-

standingly popular online tool. What could be the reasons

for this success? For one, certainly its entertaining property:

about 90 percent of participants in the Wahl-O-Mat surveys

contend that it was fun using the tool (see Figure 4.1). The

moment of surprise that occurs when the tool calculates

the results is attractive. A second reason might be its seem-

ing simplicity: the tool is easy to use and its basic mode of

operation is easy to understand. Most users are already fa-

miliar with casting votes on the Internet, as online voting

has become a very popular Web application. However, be-

hind the seemingly simple concept of the Wahl-O-Mat

there are very complex and quite costly organizational rou-

tines. In particular, the making of the propositions takes

place in a highly standardized frame, and is a complicated

process that starts several months before the tool is finally

launched.

4 For detailed information on the city-block method (“Manhattan dis-

tance” or “L1-distance”) basing on Minkowski distance, see: Bortz

(2005: 570); Backhaus et al. (2008: 404-5). Out of a methodological

perspective, other methods might seem more suitable (Klein, 2006).

But a model based on distances provides a higher degree of transparen-

cy to those using the tool.

70

Stefan Marschall - Christian K. Schmidt

4.2. The impact on the users – data analysis

As said before, millions of people have played the Wahl-

O-Mat, making it very popular. However, the exact number

of users is difficult to gauge as it is possible and very prob-

able that people play the Wahl-O-Mat several times, even

though a cookie is applied to guarantee that only one ses-

sion is counted. Accurately, one can only say that the num-

ber of user sessions reached – for example for the European

Wahl-O-Mat in 2009 – one and a half million (i.e., about

2.3 percent of the electorate), and on average ranged be-

tween two to three percent of the electorate at the state

level.

But beyond the mere quantitative perspective: what do

we know about the impact of the tool on the people who

used it? In the wake of the Wahl-O-Mat application in

2002, a face-to-face survey asked a representative sample

questions regarding the Wahl-O-Mat. However, the find-

ings did not yield profound data on impacts which could

clearly be linked to the use of the Wahl-O-Mat. In close

cooperation with a research group at the University of

Düsseldorf until 2008 and since 2008 with a research group

at the University of Siegen, the BPB has commissioned ad-

ditional survey research particularly focusing on the effects

of the Wahl-O-Mat on political behaviour.

4.3. The data base: the Wahl-O-Mat online surveys

Since the Wahl-O-Mat’s application before the 2003 Ba-

varian elections the research group has made a continuous

effort to collect data on the way voters use and are affected

by the Wahl-O-Mat (see Table 4.1). The data were gener-

ated to learn more about those using the tool and to help to

improve it. Furthermore, the findings produce evidence to

71

Chapter Four - The Impact of Voting Indicators: The Case of Wahl-O-Mat

respond to speculations regarding the power of the Wahl-

O-Mat to change the voting decision of the users.

Table 4.1 – Overview of Wahl-O-Mat online surveys

Election Survey period

Completed

interviews

Bavaria

Sep. 1 – 18, 2003

3,847

Saxony

Sep. 8 – 19, 2004

3,556

European Parliament

May 19 – June 13, 2004

12,214

Schleswig-Holstein

Jan. 12 – Feb. 20, 2005

4,315

North Rhine-West.

Apr. 19 – May 22, 2005

7,248

German Bundestag

Sep. 2 – 18, 2005

14,455

Baden-Württemberg

Feb. 15 – Mar. 26, 2006

2,990

Rhineland-Palatinate

Feb. 16 – Mar. 26, 2006

1,832

Saxony-Anhalt

Feb. 24 – Mar. 26, 2006

1,038

Berlin

Aug. 23 – Sep. 17, 2006

2,662

Bremen

Apr. 24 – May 13, 2007

1,708

Lower Saxony

Jan. 7 – 27, 2008

2,104

Hamburg

Feb. 2 – 24, 2008

932

European Parliament

May 19 – June 7, 2009

10,563

German Bundestag

Sep. 4 – 27, 2009

45,613

Source: Wahl-O-Mat Research Group, University of Siegen

For the analysis in this chapter, we draw on data from an

online survey carried out before the 2009 European elec-

tion. At some points in our argument we will integrate cor-

responding data stemming from other online Wahl-O-Mat-

surveys at national and sub-national elections or before the

European elections in 2004. The questionnaire instrument

has been rather stable over time which allows for compara-

tive analyses.

72

Stefan Marschall - Christian K. Schmidt

First, some remarks on our method: online surveys have

become a very popular (as well as comparably cheap) way

to collect data. After facing similar difficulties as mail and

phone surveys after they were applied for the first time,

online surveys have been established as a common method

of collecting data5. However, from a strict methodological

perspective, Internet-based surveys are rather contested

(see for example: Maurer and Jandura 2009; Schnell et al.,

2008: 377-86; Taddicken, 2007; Zerback et al., 2009). The

main objection addresses the problem of the representa-

tiveness of the sample. There is no guarantee that a sample

drawn by online surveying generates data which can be

generalized beyond the group of persons that were inter-

viewed. Moreover, there is little control over who is really

filling out the questionnaire, due to the anonymity of online

communication.

However, for our research there were some profound

reasons to use web-based surveys: the questionnaire is di-

rected at individuals who should definitively have played

the Wahl-O-Mat. The best way to reliably recruit people

who can evaluate their experiences with the Wahl-O-Mat is

to get hold of them online right after they have played the

tool. In terms of representativeness, we were not interested

in creating a sample representative of the German elector-

ate. Nor were we looking for a sample that adequately re-

flects the composition of the online community. The

members of the target population we were interested in

were only those who have used the tool. Thus, we had to

generate a sample representative of the people using the

Wahl-O-Mat. For this purpose, the best choice was the

5 Schonlau et al. (2001: iii); for further information on methodology and

types of online surveys, see: Couper (2000); Welker et al. (2005).

73

Chapter Four - The Impact of Voting Indicators: The Case of Wahl-O-Mat

method of online surveying being directly connected to the

Wahl-O-Mat.

Technically, we recruited the sample we will be referring

to in this paper (“European Parliament 2009”) by asking

users directly after their Wahl-O-Mat sessions whether they

would be willing to fill in an online questionnaire; in the

moment the Internet users were about to leave the Wahl-

O-Mat-site, a window popped up inviting them to help to

improve the tool by taking part in a survey (“on exit”).

The recruitment of the sample was based on random

procedure. In order to solve the problem of increasingly

popular pop-up blockers within the last years we used a so

called “layer pop-up” procedure. The layer pop-up window

did not open after every session to reduce the probability

that people who (used) the tool several times would be

asked to participate more than once. Every tenth person

using the Wahl-O-Mat on one of the BPB-servers (leaving

out other servers of, for example, media partners) was

asked to take part in the survey6. Cookies ensured that us-

ers who had already been confronted with the invitation

were not invited to take part again.

Altogether, 73.742 individuals were invited to fill in the

questionnaire (see Table 4.2). The response rate ranged

about 14.3 percent. In the end, we drew on 10,563 com-

pleted questionnaires. Although the response rate of web-

based surveys is difficult to assess7, compared to other sur-

6 The generation of the sample is based on the Nth Visitor Methodolo-

gy, a technique developed by U.S. marketing researchers in order to

prevent self-selection bias (Pfleiderer, 2003: 385-7).

7 Theobald (2003: 203) and Schonlau et al. (2001: 81ff.) provide an

overview on difficulties of rating different response rates.

74

Stefan Marschall - Christian K. Schmidt

veys of this type the quota seems to be absolutely suffi-

cient8.

Table 4.2 – Basic information on the 2009

European Parliament election Wahl-O-Mat survey

Survey period

May 19 to June 7, 2009

Invited to participate

73,742

Completed interviews/participants

10,563

Response rate

14.3 %

Participants entitled to vote

9,966

Source: Wahl-O-Mat Research Group, University of Siegen

Still, the level of representativeness of the sample for the

population of Wahl-O-Mat users is difficult to assess. Al-

though we had a huge number of completed question-

naires, we were not sufficiently confident because – as it is

well known – an increased sample size does not automati-

cally equate to increased representativeness. The same goes

for relatively high response rates: although they definitely

help to provide for representativeness, they cannot give

certainty.

However, other aspects made us confident that our re-

sults have descriptive and possibly power for those using

the Wahl-O-Mat. First, we compared the survey’s distribu-

tions on some variables with other data to check for repre-

sentativeness. As mentioned above, we did not expect our

sample to be completely representative for the population

in Germany. Still, in some respects we assumed a signifi-

cant correspondence (e.g., between the territorial distribu-

8 Findings of El-Menouar and Blasius (2005: 79) and Theobald (2003:

207-8) are very helpful in order to assess response rates of different

types of surveys.

75

Chapter Four - The Impact of Voting Indicators: The Case of Wahl-O-Mat

tion of the sample on the one hand and of the population

on the other), for there is no strong plausibility for a terri-

torially skewed distribution of the Wahl-O-Mat-population.

This indeed was roughly confirmed (see Table 4.3).

Table 4.3 – Territorial distribution of survey sample and population

State / Bundesland Population Interviewed

Baden-Württemberg

13.1 %

11.2 %

Bavaria

15.3 %

12.2 %

Berlin

4.2 %

8.0 %

Brandenburg

3.1 %

2.8 %

Bremen

0.8 %

1.5 %

Hamburg

2.2 %

2.7 %

Hesse

7.4 %

5.6 %

Mecklenburg-Western Pomerania

2.0 %

1.1 %

Lower Saxony

9.7 %

9.1 %

North Rhine-Westphalia

21.9 %

30.9 %

Rhineland-Palatinate

4.9 %

3.5 %

Saarland

1.2 %

1.1 %

Saxony

5.1 %

4.1 %

Saxony-Anhalt

2.9 %

1.6 %

Schleswig-Holstein

3.5 %

2.7 %

Thuringia

2.8 %

2.0 %

Average absolute deviation: 1.8 percentage points

Sources: <http://www.statistik-portal.de/Statistik-

Portal/en/en_jb01_jahrtab1.asp>, accessed 11/15/2009;

Web survey Wahl-O-Mat EP elections 2009, n=10,563

Second, the gender distribution served as a point of con-

trol. We identified a high correspondence between the gen-

76

Stefan Marschall - Christian K. Schmidt

der distribution within the survey sample and the gender

distribution in the German Internet population referring to

the results of a representative survey conducted by AGOF

(Arbeitsgemeinschaft Online Forschung) 2009. About 51% of in-

terviewees were male, which corresponds approximately to

the proportion of males within the German online com-

munity.

Third, we gained confidence in the representativeness of

our sample and the reliability of the method by comparing

the results of the 2009 survey with the findings in surveys

we had previously conducted on the Wahl-O-Mat. Indeed,

some response patterns have been extremely stable over

the time. For example, users agreed with the item “It was

fun playing the Wahl-O-Mat” about 90 percent of the time

in all surveys so far (see Figure 4.1).

Figure 4.1 – Respondents declaring that

“it was fun playing the Wahl-O-Mat” (in percent)

Source: Wahl-O-Mat Research Group, University of Siegen

Besides the points mentioned above, we additionally

checked for further consistencies within the data (i.e., type

of Internet access, response time, etc.). Although an error

probability remains difficult to assess, all indicators have

encouraged us to assume that our sample’s data provide

77

Chapter Four - The Impact of Voting Indicators: The Case of Wahl-O-Mat

valid information about our target population, i.e., those

using the Wahl-O-Mat.

4.4. The Wahl-O-Mat population – findings

What do we know about those playing the Wahl-O-Mat?

What does a typical Wahl-O-Mat-user look like? First, the

Wahl-O-Mat community is rather young (see Table 4.4).

About 48 percent are younger than 30 years. Evidently, this

distribution does not correspond with the demographic

characteristics of the German society: those under 30 years

in the German population account for approximately 20

percent. This finding is unsurprising, considering that the

age distribution within the online population does not

match the age distribution of the society at all. However,

the Wahl-O-Mat-demographics do not perfectly match the

distribution among the online community, either. The users

of the Wahl-O-Mat are on average younger than the aver-

age Internet user. Within the online population, only about

31 percent are younger than 30 years. Strongly underrepre-

sented in the Wahl-O-Mat community is the segment of

the so-called “silver surfers”, people of 60 years and older:

about six percent in the Wahl-O-Mat population as op-

posed to about 11 percent in the German online popula-

tion.

Second, those using the Wahl-O-Mat have a rather high

formal educational background: three quarters of our sam-

ple consist of persons with baccalaureate/A-Level or an

academic degree. Heavily underrepresented are people with

a basic formal education only. They constitute the smallest

segment in the Wahl-O-Mat community.

78

Stefan Marschall - Christian K. Schmidt

Table 4.4 – Age groups, aged 14 and above (in percent)

Age (in years) <20 20-29 30-39 40-49 50-59 59<

Wahl-O-Mat Users

12.7

34.7

20.2

16.6

10.0

5.9

Online population

11.5

19.4

19.3

23.3

15.1

11.4

General population

7.8

13.5

14.4

19.3

15.5

29.4

Sources: Web survey European Parliaments elections 2009 (n=10,563),

Wahl-O-Mat Research Group, University of Siegen;

AGOF: Internet Facts, Berichtsband zu Internet facts 2009-II,

<http://www.agof.de/berichtsbandif2009ii.download.848e0b18fb27fa

58bbe 4541db0354a5c.pdf>; accessed 9/30/2009.

We can also read from the data reported in Figure 4.2

that, on average, Internet users have a higher formal educa-

tional background than the German population (AGOF

2009: 7; Gerhards et al., 2009, Schmitt-Beck et al., 2005). In-

terestingly, Wahl-O-Mat users have attained an even better

formal education than the average online population. This

might tie in with another typical feature of the Wahl-O-Mat

population, the extent of political interest and activities.

For our third basic finding is that Wahl-O-Mat users are

highly interested in politics and are more politically active

than the population on average. More than three quarters

of those interviewed considered themselves politically in-

terested. About 60 percent claimed to frequently discuss

political questions. Those organized in parties are overrep-

resented in the sample: in Germany, no more than about

two percent of citizens are registered members of parties,

whereas in the Wahl-O-Mat community 5.6 percent dis-

closed that they were members of a political party. Only a

small percentage of respondents contended that they will

not go to the ballot boxes although they are by law entitled

to vote. About 90 percent of the Wahl-O-Mat population

planned to go voting, whereas in fact only 43.3 percent of

79

Chapter Four - The Impact of Voting Indicators: The Case of Wahl-O-Mat

the German electorate took advantage of their right to vote

in the European elections of 2009.

Figure 4.2 – Formal education (in percent)

Sources: Web survey Wahl-O-Mat EP elections 2009 (n=10,563),

Wahl-O-Mat Research Group, University of Siegen;

AGOF Internet facts 2009-II

There is a further dimension of political interest and

commitment of the Wahl-O-Mat users that distinguishes

them from the overall population. At one point in the ques-

tionnaire, interviewees could indicate whether they had a

clear party preference. About 85 percent (see Figure 4.3)

contended that they indeed had a party identification,

whereas in the German population the group of undecided

voters has been growing continuously. Only a small

amount, approximately 10 percent, of those interviewed

might reconsider their vote due to the Wahl-O-Mat, show-

ing that the party preferences of Wahl-O-Mat users are in-

deed comparably robust.

80

Stefan Marschall - Christian K. Schmidt

To sum it up: the typical user of the Wahl-O-Mat is

young, highly educated, politically interested, and actively

participating in politics.

Figure 4.3 - Extent of political interest and participation

of Wahl-O-Mat users (in percent)

Source: Web survey Wahl-O-Mat European Parliament elections 2009

(n=10,563; n= 9,966 for “plan to go voting”)

Wahl-O-Mat Research Group, University of Siegen

4.5. Mobilizing impacts of the Wahl-O-Mat

Having learned more about the people using the Wahl-O-

Mat, we now turn to the mobilizing effects the tool has on

users. Several items in the questionnaire directly addressed

aspects of political mobilization.

By asking about the implications on political participa-

tion, we analyzed whether there is a connection between

online communication and offline participation. At the out-

set, playing the tool itself is a form of political participation:

81

Chapter Four - The Impact of Voting Indicators: The Case of Wahl-O-Mat

the users take positions and even vote on relevant political

questions, leaving data traces on the Internet. Yet, besides

mere online activities, we are interested in the mobilizing

power of the tool, which goes beyond the virtual realm, be-

coming manifest within the “real” world, that is, in the

world of offline participation.

We directed our research on offline forms of involve-

ment (e.g., discussing politics with others, searching for fur-

ther information, and voting). The concept of mobilization

on which we based our analysis is rather broad. Mobiliza-

tion in this sense is not only the capacity of parties to rally

their members. Our understanding of the concept includes

categories like activation and knowledge gain (Schmitt-Beck

and Farrell, 2002: 13).

Firstly, we asked the users whether playing the tool stimu-

lates them to talk about the result with others (friends and

family). Here, we tried to establish whether playing the

Wahl-O-Mat contributes to a communicative and social

dimension of political action. Furthermore, this form of

political participation might have a multiplying effect: by

talking about the tool, the elections and politics in general

might emerge as a topic of day-to-day discussion, constitut-

ing a “two step flow” of communication (Katz and Lazars-

feld, 1955). However, we did not ask exactly how and in

which context the users talked about the Wahl-O-Mat (e.g.

whether they were going to deliberate the political ques-

tions raised by the application, or whether they would talk

about the tool itself).

Secondly, the users of the Wahl-O-Mat were asked if

playing the tool motivated them to gather further informa-

tion about the election and the parties taking part in it.

With this item, we were looking at how the tool contrib-

uted to a change in the information seeking behaviour, ad-

dressing a “cognitive dimension” of participation. The

82

Stefan Marschall - Christian K. Schmidt

search for more information can serve to increase one’s

competences in understanding politics by affecting the ex-

tent and quality of individuals’ political activities.

Thirdly, the users were asked whether playing the tool has

mobilized them to go to the polls – the “activity dimen-

sion”. As mentioned at the beginning, the decrease in voter

turnout was (one) central impulse for the BPB to launch

the tool. Therefore, the question whether the Wahl-O-Mat

encourages people to vote has received a lot of attention.

With this item, we looked for a specific and isolated Wahl-

O-Mat effect by asking whether the Wahl-O-Mat has moti-

vated people to go to the polls, even if they had not in-

tended to do so before playing the tool.

All these items comprise as much an ex post as an ex ante

perspective. They are ex post in asking users directly after

their Wahl-O-Mat-session about the effects playing the tool

could have. The ex ante perspective refers to the users being

asked to anticipate the mobilizing effects. However, at this

point of time this is just a subjective prediction. We do not

know for sure whether those saying that the Wahl-O-Mat

has motivated them to collect further political information

were really going to do so. We cannot take it for granted

that the users who claimed to vote or to talk with friends

and family about the Wahl-O-Mat results were in fact doing

so. The tendency to answer questions like those referred to

above as potentially socially desirable can produce biases

(Schnell et al., 2008: 355-6). Moreover, people could also be

incapable of reliably predicting their own future behaviour,

due to intervening factors and changing circumstances that

they cannot take into account when confronted with the

questions.

To ask the respondents at a second point of time whether

the Wahl-O-Mat did indeed have the mobilizing effects

they had assumed it would have (e.g., right after polling

83

Chapter Four - The Impact of Voting Indicators: The Case of Wahl-O-Mat

day), would increase validity. However, the design chosen

for the survey does not allow for this kind of ex-post con-

trol and/or panel analysis. At this stage we must be content

with the data provided, keeping in mind that we are not

looking at de facto mobilization effects, but at subjectively

predicted ones.

Let us turn to the results of the surveys. First we will look

at the univariate distribution. Then we will refer to some

cross tabulations qualifying some of the results.

Figure 4.4 – Mobilization (in percent)

Source: Web survey Wahl-O-Mat EP elections 2009 (n = 10,563),

Wahl-O-Mat Research Group, University of Siegen

The first striking finding is a huge variation between the

different items of mobilization (see Figure 4.4). Nearly 70

percent of the sample users contended that they were going

to talk about the tool with others; almost 60 percent said

that they had been stimulated by the Wahl-O-Mat to seek

n = 9,966

persons entitled

to vote

84

Stefan Marschall - Christian K. Schmidt

additional political information, whereas only about 11 per-

cent were motivated to go to the ballot boxes. It is neces-

sary to take a closer look at these findings.

Talking with others about the results of the Wahl-O-Mat

must be considered a rather low scale form of participation.

It is not very surprising that those users who usually engage

in political discussions discussed the Wahl-O-Mat and its

results (74.8 percent, see Table 4.5). But out of those who

do not discuss politics frequently 62.5 percent (i.e., almost

two thirds), were motivated to talk about the tool, too. The

social and communicative effect was not restricted to the

“talking people”.

Table 4.5 – “I will probably talk about the result with friends and family” (%)

“I discuss on political issues frequently”

Yes No DK Total

Yes

74.8

62.5

63.8

69.9

No

16.0

25.4

16.3

19.0

Don’t know

9.2

12.1

19.9

11.1

Cramer’s V = 0.108, p < 0.001

Source: Web survey European Parliament elections 2009 (n = 10,563),

Wahl-O-Mat Research Group, University of Siegen

Almost 60 percent claimed that the Wahl-O-Mat had

stimulated them to collect further political information. Of

course, there is a high association between political interest

and the motivation to get more information on politics.

Many of those using the Wahl-O-Mat expressed their po-

litical interest already when they played the Wahl-O-Mat in

order to collect information about the parties and their po-

sitions. 59.6 percent said that this had been the dominant

85

Chapter Four - The Impact of Voting Indicators: The Case of Wahl-O-Mat

motivation for them to go to the Wahl-O-Mat site. Regard-

ing only the users who said the most important reason to

use the Wahl-O-Mat was to get more information on the

political positions of the parties, the value of the item “mo-

tivated me to collect further political information” increases

to 76.7 percent. In comparison: just 48.3 percent of the

“only curious” sought political information because of

playing the Wahl-O-Mat.

Taking a closer look at the data, we seem to identify one

variable being of explanatory power for this dimension of

mobilization: gender. In Table 4.6, we have displayed the

distribution controlling for gender in our sample on the

item: “The Wahl-O-Mat has motivated me to collect fur-

ther political information”. We see a small but significant

gap between the two groups: whereas only 54.9 percent of

the male respondents were motivated to seek further in-

formation on politics, 64.5 percent of female users were

moved to inform themselves further.

Table 4.6 – “The Wahl-O-Mat has motivated me to collect further

political information” (in percent)

Gender

Male Female Total

Yes

54.9

64.5

59.6

No

29.6

21.3

25.5

Don’t know

15.5

14.2

14.9

Cramer’s V = 0.104, p < 0.001

Source: Web survey Wahl-O-Mat EP elections 2009 (n = 10,563),

Wahl-O-Mat Research Group, University of Siegen

86

Stefan Marschall - Christian K. Schmidt

Yet the differences in the Wahl-O-Mat’s mobilization

power between male and female users are closely linked to

the finding that generally more male participants in the

sample viewed themselves as politically interested than fe-

male participants (86.2 as to 66.4 percent). Women seem to

consider themselves being in a subjectively defined need

for more political information.

The number of those users whom the Wahl-O-Mat moti-

vates to vote is small simply because the third item focused

on those who had not intended to vote before they started

playing the Wahl-O-Mat. This item is a combination of two

variables: (1) Did you plan to vote before you played the

Wahl-O-Mat, (2) Does the Wahl-O-Mat make you vote? As

mentioned above, the typical Wahl-O-Mat user is politically

highly interested. Thus, the section of the sample that was

open for this form of mobilization was rather small. In-

deed, of those who considered themselves not to be politi-

cally interested, almost 20 percent mentioned this special

mobilizing effect (see Table 4.7).

Table 4.7 – “Actually I did not want to vote.

The Wahl-O-Mat has motivated me to go voting” (in percent)

Interested in Politics

Yes No DK Total

Yes

8.8

19.8

14.1

10.8

No

84.0

68.3

70.0

80.5

Don’t know

7.3

12.0

15.9

8.7

Cramer’s V = 0.119, p < 0.001

Source: Web survey Wahl-O-Mat European Parliament elections 2005

(n=9,966; respondents not entitled to vote excluded),

Wahl-O-Mat Research Group, University of Siegen

87

Chapter Four - The Impact of Voting Indicators: The Case of Wahl-O-Mat

There are some other striking findings concerning the

variable “going to vote though did not intend to do so”.

There seems to be a strong association between the educa-

tional background and the power of the Wahl-O-Mat to

push people to go to the elections (see Figure 4.5). The

higher the degree of formal education is, the lower ranges

the capacity of the Wahl-O-Mat to mobilize people to vote.

Yet this finding must be interpreted carefully, too. The im-

portant point is that the levels of political interest and par-

ticipation are augmented by the educational degree. This

means that among those with a university degree there are

fewer who the Wahl-O-Mat could mobilize because they

had already been politically active before playing the tool.

Figure 4.5 – Formal education attainment (mobilization to vote, %)

Source: Web survey Wahl-O-Mat European Parliament elections 2009

(n = 9,966; respondents not entitled to vote excluded),

Wahl-O-Mat Research Group, University of Siegen

88

Stefan Marschall - Christian K. Schmidt

4.6. Discussion: deep impact?

What can be said about the impact of the Wahl-O-Mat?

What are the major findings? We found strong evidence

that an online application like the Wahl-O-Mat has the

power to mobilize individual offline political participation.

The tool stimulates activities that go far beyond just using

an Internet device. Moreover, in the wake of using the net

application we observe effects in different forms of offline

political participation. Users talked about the results the

tool has processed; users were stimulated to look for fur-

ther political information; users were motivated to go to

the ballot boxes although they had not planned to do so

before they started playing the Wahl-O-Mat.

Still, for methodological reasons we had to rely on the us-

ers’ perceptions and their subjective ex-ante guesses about

what effects the tool might have on their political participa-

tion. Being restricted to data generated by an on-exit survey

right after the use of the tool, we had no opportunity to re-

visit the respondents at a later point of time in order to ask

whether the expected effects had indeed become reality.

Nor could we apply other methods (e.g., participant obser-

vation) to precisely register manifest effects on participa-

tion. Therefore, uncertainty remains about whether the

predicted consequences really did take place. Nevertheless,

because the numbers were very high, we expect that, if not

the entire predicted amount, then at least a reasonable pro-

portion of the expected mobilization really did occur.

Our findings indicate that Internet activities are not insu-

lated, but that there is a “link” between visiting political

websites and offline participation. The “first link” in this

chain refers to the group of those using political websites.

We found that the typical Wahl-O-Mat user is neither an

average citizen nor an average Internet surfer. But the peo-

89

Chapter Four - The Impact of Voting Indicators: The Case of Wahl-O-Mat

ple using the application have a high educational back-

ground and represent the segment of politically active citi-

zens. Wahl-O-Mat users consider themselves interested in

politics; they participate in parties on average more than

other citizens and frequently take part in political discus-

sions. Thus, many of those playing the tool were already

politically active before coming into contact with the Wahl-

O-Mat.

The “second link” occurs after the users have played the

tool. After the Wahl-O-Mat sessions, the quality and extent

of participation changes within a large group of users

merely because of this usage. This is a clear media effect, a

change which can unmistakably traced back to using a spe-

cific media application, here an Internet tool (for similar

findings, see: Emmer and Vowe, 2004: 204-6).

But who is being mobilized? Does the tool mobilize the

mobilized, or does it really make a difference by increasing

political participation beyond what could normally be ob-

served? At first glance, it might seem that the Wahl-O-Mat

mobilizes those who participate intensively anyway. People

who are usually inclined to talk about politics are going to

talk about the Wahl-O-Mat and its results, too. The Inter-

net motivates those who usually seek political information

(for example, by using the Wahl-O-Mat!) to continue with

this political engagement.

Beyond the mobilization of the mobilized, in our analysis

we could identify effects on those who consider themselves

not politically interested, among those who usually do not

participate in politics. 14 percent of the interviewees are not

at all politically involved. Although comparably small, this

group does exist and is also partly mobilized by the Wahl-

O-Mat. Thus, the Wahl-O-Mat is able to move people who

are not by nature prone to participation.

90

Stefan Marschall - Christian K. Schmidt

In one item we particularly looked for a significant

change of behaviour by addressing people who planned to

refrain from one specific but salient form of political par-

ticipation before they used the Wahl-O-Mat: taking part in

elections. Here we found a small, yet remarkable number of

people who were motivated by the tool (and only by the

tool) to take advantage of their right to vote. Considering

the popularity of the application and the high number of

people using the Wahl-O-Mat, this effect may influence

about hundreds of thousands of people in the federal elec-

tion – this being a very tentative guess.

To sum it up: Tools such as the Wahl-O-Mat have be-

come increasingly popular. Voting indicators have emerged

as indispensable elements in pre-election periods. To some

degree they will make a difference, they will have a chang-

ing impact on the political and democratic culture of a soci-

ety – judged from our findings: for the better.

CHAPTER FIVE

More than toys? A first assessment of voting

advice applications in Switzerland1

ANDREAS LADNER, GABRIELA FELDER, JAN FIVAZ

To what extent are Voting Advice Applications (VAAs)

more than toys and should political scientists be held ac-

countable for the VAAs they produce? A toy is basically an

object to play with, but toys are also important tools for

learning about the real world and promoting the process of

socialisation. If VAAs are toys they are meant to playfully

attract people to politics, provide them with information,

increase their interest in politics and motivate them to par-

ticipate in elections. If they are more than toys they addi-

tionally have a direct impact on the votes of their users and

therefore on the outcome of elections. In this sense it is no

longer the aspect of ‘learning by playing’ but much more

the aspect of being an important element in the course of

elections which has to be addressed. And: If we have to

admit that VAAs have an impact on the outcome of elec-

tions then the second question becomes important. If

VAAs are to be taken seriously to what extent can their

providers be held accountable? Should they only be ac-

countable for the quality of the tool itself or also for a pos-

sible influence on the outcome of elections? Can a clear

distinction be made between offering a new form of sup-

1 An earlier version of this paper was presented at the “Voting Advice

Applications (VAAs): between charlatanism and political science” con-

ference at the University of Antwerp, May 16, 2008.

92

Andreas Ladner - Gabriela Felder - Jan Fivaz

port for decision-making and influencing electoral behav-

iour?

Based on the experience with the increasingly popular

Swiss VAA smartvote2 and the results of a major research

project analysing the use and impact of smartvote on the

Swiss national elections in 20073 we shall – although tenta-

tively at this stage – try to answer these two questions.

For a better understanding of the functioning and the

importance of smartvote (section 5.2) we will start by looking

at some characteristics of the Swiss electoral system (sec-

tion 5.1). In sections 5.3 and 5.4 we will present empirical

evidence about the use of smartvote and the role and the im-

portance attached to it by voters and candidates. Section

5.5 will focus on the accountability question and the possi-

bilities and limits of VAAs within the Swiss legal frame-

work. The final section 5.6 offers a short conclusion and an

outlook on further developments and challenges likely to

occur in the years to come.

5.1. Elections and the Electoral System in Switzerland

Design and set-up of a VAA as well as its use by parties,

candidates and voters depend largely on country-specific

characteristics of the electoral system and the way citizens

elect parties or candidates. Both the electoral system and

the low turnout in elections make VAAs in Switzerland es-

pecially useful and important.

Politics in Switzerland take place in a very fragmented so-

cial context. The country is divided into 26 cantons, which

2 <http://www.smartvote.ch>.

3 The research project was funded by the Swiss National Science Foun-

dation. It is part of a large research programme called “Challenges to

Democracy in the 21st Century” (NCCR Democracy).

93

Chapter Five - More than toys? A first assessment of VAAs in Switzerland

also form the electoral constituencies for the elections of

the national parliament. The Swiss parliament consists of

two symmetric and non-congruent chambers (Lijphart,

1999): the National Council (Nationalrat) and the Council of

States (Ständerat). The National Council has 200 seats and is

elected by means of a proportional system; the Council of

States has 46 seats and is elected by a majority system4.

Thus elections for the National Council are generally con-

sidered as more party-oriented and the elections for the

Council of States as candidate-oriented.

The seats for the National Council are assigned to the

cantons according to their population size: the six smallest

cantons have only one seat whereas the canton of Zurich,

the largest canton, has 34 seats. Accordingly, the number of

candidates running for office varies between one candidate

in the canton of Uri and 804 in the canton of Zurich

(Fivaz, 2007; Bundesamt für Statistik, 2007). The cantons

differ also in various other aspects: language, religion and

economic structure. Subsequently, cantonal party systems

differ widely for example with regard to the number of par-

ties and the degree of party competition (Ladner 2004a;

2004b).

A further aspect of the social and political heterogeneity

of Switzerland is the fragmentation of the political parties

(Ladner, 2002). Switzerland has a large number of parties

with a relatively low share of votes, parties are decentralised

and the cantonal and local sections have far-reaching

autonomy. Furthermore, it is not unusual that there are di-

verse political positions within a single party. Even individ-

4 There are some exceptions to these rules: In cantons with just one

seat in the National Council the effects of PR disappear and the canton

of Jura uses the proportional counting procedure for the election of the

Council of States as well.

94

Andreas Ladner - Gabriela Felder - Jan Fivaz

ual candidates may take autonomous positions (see table

5.10) and resist the dictate of their party leaders.

While electing their members of parliament Swiss voters

have the possibility to express their specific preferences for

parties as well as for single candidates. First, every voter has

as many votes as his constituency has seats (e.g., in the can-

ton of Uri with one seat, voters have one vote and in the

canton of Zurich with 34 seats they have 34 votes). Sec-

ondly, voters can split their votes among different parties

(e.g., in the canton of Zurich a voter can give four votes to

party A, ten to party B and 20 to party C). Thirdly, voters

can support their favourite candidates by giving them two

votes instead of one (so-called cumulative voting; e.g., in

the canton of Zurich a voter could vote for 17 candidates

with two votes for each). These rules make it possible to

compose a customized ballot according to one’s personal

political preferences.

Due to the fragmentation of the political and the party

system Swiss voters can choose among a big number of

parties and political positions, and quite often it is rather

difficult to get to know all parties and candidates (particu-

larly in a canton like Zurich with over 800 candidates).

Compared to voters confronted with a two-party system it

is definitely more time-consuming for Swiss voters to

gather the necessary information about parties and candi-

dates. Nevertheless, Swiss voters seem to appreciate these

possibilities increasingly. The share of swing voters has in-

creased in the last years (Linder, 2005) as well as the share

of those using the possibilities offered by the electoral sys-

tem to compose their customized ballots according to their

individual preferences (Burger, 2001). Here, candidate-

based VAAs like smartvote step in and offer the much

needed information for choosing appropriate parties and

candidates.

95

Chapter Five - More than toys? A first assessment of VAAs in Switzerland

Despite the far-reaching possibilities to express one’s

preferences, electoral turnout in Switzerland is very low

compared to other countries and this is not an entirely new

phenomenon. Since 1975 electoral participation has never

been higher than 50 percent. The lowest score up to now

was in the 1995 elections when only 42.2 percent of those

entitled to vote participated. Since then turnout has in-

creased again: 43.3 percent in 1999, 45.2 percent in 2003

and 48.3 percent in 20075. In contrast to countries with

turnout rates around 80 percent, a large proportion of

Swiss voters are waiting to be mobilized, which is a wel-

come challenge for VAAs trying to increase political par-

ticipation.

5.2. Differences between smartvote and other VAAs

There are two major VAAs in Switzerland. The smaller

one, which is called Politarena, is based on the concept of

the pioneer platform StemWijzer, very much like the Ger-

man Wahl-O-Mat. The bigger one, smartvote, takes a different

approach which adapts much better to the specific charac-

teristics of the Swiss electoral system and the needs of the

voters. The concept of smartvote has been the basis for other

applications such as Politikkabine, Koimipasva and Holyrood.

Compared to other VAAs smartvote is more comprehen-

sive as regards its additional features as well as its extensi-

bility. The main differences between smartvote and its com-

petitors are the following (for a better discussion, see: Fivaz

and Schwarz, 2007: 6f):

- smartvote is capable of managing multiple elections with

5 <http://www.politik-stat.ch/nrw2007CHwb_de.html>; accessed

27/11/2009.

96

Andreas Ladner - Gabriela Felder - Jan Fivaz

overlapping constituencies at the same time (e.g., one

national, one cantonal and two local elections).

- smartvote calculates voting recommendations according

to the electoral system and constituency (electoral dis-

trict)6 at the level of single candidates as well as at the

level of lists/parties.

- The smartvote-questionnaire – which contains more than

70 questions – is more than twice as long as question-

naires used by other tools. Hence the recommendation

is based on more empirical data and therefore more reli-

able.

- Besides Kieskompas, smartvote is the only VAA which in-

cludes additional visual analytical tools like the smartspi-

der and the smartmap graphs (see figure 5.1 and figure

5.2).

- Finally, time series analyses are possible as all the data of

past elections are stored.

Figure 5.1 – smartspiders of Liberals, Christian Democrats and Green Party

Source: Neue Zürcher Zeitung, 24th October 2007

6 StemWijzer for instance provides one recommendation for the whole

election. In Switzerland not every party necessarily runs for election in

every constituency and local and regional party sections might vary in

their political positions, hence a meaningful voting recommendation

has to account for these specific circumstances.

97

Chapter Five - More than toys? A first assessment of VAAs in Switzerland

Figure 5.2 – smartmap of Swiss parties in the National Council

Source: Neue Zürcher Zeitung, 24th October 2007

5.3. The Use of smartvote

When smartvote was first presented to the voters in 2003 a

modest number of 255.000 ‘voting recommendations’7

were made, while Politarena reached 135,000 users. Since

then, VAAs have become increasingly popular. During the

run-up to the elections for the Swiss parliament in October

7 In Switzerland this is a prevalent term which may be different in other

countries where VAAs have come into use.

98

Andreas Ladner - Gabriela Felder - Jan Fivaz

2007 smartvote issued about 963.000 recommendations.

Compared to 2003 the use of smartvote had increased almost

fourfold in 2007.

The increasing use of VAAs can certainly be explained by

technical progress and the increase of Internet access. In

2006 over 75 percent of the Swiss population had access to

the Internet8. Besides the high rate of Internet access there

are additional factors that are fostering the popularity of

VAAs. Political parties are facing severe challenges: Within

the last 20 to 30 years traditional ties between voters and

parties are loosening (Dalton and Wattenberg, 2000; Wal-

grave et al., 2008a), the number of party members is de-

creasing and the volatility rate and the number of swing

votes is rising. Dalton, for example, draws quite a pessimis-

tic picture of representative, party-centred democracies

with more and more citizens growing distrustful of politi-

cians and disillusioned about the functioning of the democ-

ratic processes (Dalton, 2002; 2007). Although it is still an

open question to which degree this pessimistic picture of

today’s representative democracies meets reality we assume

that these developments – at least in their tendency – foster

the use of VAAs, which are offering a customized and

transparent new form of decision-making beyond the usual

ways of selecting candidates and parties.

Some further figures about the use of smartvote on both

sides – the one of the voters as well as the one of the par-

ties and candidates running for office – will document the

growing importance of such tools.

8 See <http://www.bfs.admin.ch/bfs/portal/de/index/themen/16/

04/key/approche_ globale.tables.30106.html>; accessed 28/04/2008.

99

Chapter Five - More than toys? A first assessment of VAAs in Switzerland

5.3.1. Use by Candidates

Since smartvote does not code party positions but asks the

candidates to position themselves and also takes into ac-

count the positions of every single candidate, the participa-

tion of all parties and candidates is an essential precondi-

tion for the additional value smartvote offers to the voters.

Unless all relevant candidates are in the database the addi-

tional value for the voters is rather low. To what extent do

the candidates answer the smartvote questionnaire? And

what are the incentives to take part?

The percentage of candidates answering the 73 questions

is a first indication for the seriousness of the VAA smartvote.

Table 5.1 highlights a sharp increase of interest in smartvote

in the National Council elections of 2007. In 2003, only

about 50 percent of the almost 3000 candidates participated

and answered the questions. Four years later, about 85 per-

cent of the 3100 candidates took part in smartvote and an-

swered the questions. If we look at the candidates elected

in the course of the elections, the figures are even more

impressive. In 2003 about 70 percent of the candidates

elected participated in smartvote, and in 2007 more than 90

percent did so. This extraordinary coverage also holds for

elections at lower level, which are also depicted in table 5.1.

Such high percentages make it possible to calculate and is-

sue meaningful voting recommendations for the public.

Thanks to media partnerships with relevant Swiss media

(from SF DRS, NZZ Online to 20Minuten)9 smartvote man-

aged to extend its reach far beyond the Internet commu-

nity. The media published articles and portrayed the candi-

dates with the aid of the political profiles generated by

smartvote; they broadcast telecasts or radio transmissions re-

9 See all media partners <http://www.smartvote.ch/side_menu/

partner/partners.php?who=v>; accessed 28/04/2008.

100

Andreas Ladner - Gabriela Felder - Jan Fivaz

ferring to the VAA smartvote; and the print media used the

visual analytical tools such as the smartspider (see figure 5.1).

Media and candidates depend on each other. On the one

hand, candidates have a greater motivation to publish their

political preferences in the VAA when they know that large

media partners will spread their political profiles also in the

press and, on the other hand, the media themselves have an

obvious interest in having a well-populated database at their

disposal.

Table 5.1 – smartvote-participation by candidates, 2003-2008

Elections Participation by

candidates (%)

Participation of

elected MPs (%)

Swiss parliament 2003

50.3

69.5

Swiss parliament 2007

85.3

93.5

Regional parliaments

Canton of Thurgau 2004

62.9

77.7

Canton of St. Gallen 2004

72.9

78.9

Voting Advice Applications in Europe: The State of the Art

Figures