3 The design of the study: the distinguishing characteristics of our approach
MARTIN DOLEZAL
This study deals with changing conflict structures in West European societies and their mobilization by political parties. In general, there are two basic approaches to handling such a question: the first one con- centrates on the changing relationships between political actors and often resorts to concepts of network analysis (Laumann and Knoke 1987; Laumann and Pappi 1976; Knoke et al. 1996; Scott 2000; Wassermann and Faust 1999). In this book, we follow the second approach and analyze issue-positions of parties as well as of voters, since we are especially interested in the thematic basis of political conflicts. This is not only the standard approach in political science (e.g. Kitschelt 1994, 1995) but has recently become even more impor- tant because of the growing significance of issue-based voting behaviour (Downs 1957; Key 1966; Budge and Farlie 1983a; Franklin 1985; Aardal and van Wijnen 2005). We expect the parties, the main political actors in West European democracies, to select the issues they articulate in party competition as well as their positions strategically. But we also look at the other side of this competition where we are interested in the changing issue-positions of the voters. However, contrary to pure rational choice or individualistic approaches, we combine the issue-based approach with a structural perspective which is focused on the political attitudes of groups. According to our point of view, membership in social groups still constitutes an important basis for the development of issue- preferences. As explained in the two introductory chapters, we conceive of contemporary cleavage structures as being thematically determined by issues linked to globalization and as being structurally rooted in social groups of winners and losers of globalization.
In this chapter, we present the design of our study. First, we briefly review how we selected the countries and elections we cover. Then we explain the kind of methods we used in the analyses of party competition for which we always distinguish two sides: the demand side, i.e. the preferences of the voters, on the one hand, and the supply side, i.e. the
53
54 Martin Dolezal
parties’ programmatic statements, on the other hand. Throughout the subsequent chapters, both aspects of electoral competition will be addressed primarily with spatial concepts resulting in standardized figures that will allow the reader to follow the effect of the new cleavage on the configuration of the national political space in the countries compared. The dimensionality of the space, which is the number of basic lines of conflict, and the content or ‘nature’ of these conflicts, are crucial points in our analysis. Note that we do not start with a priori assumptions of the space’s structure but regard the unfolding of its configuration and the substantive content of its dimensions as a major step of the analysis. As our approach is novel especially concerning the supply side analyses, we shall introduce this part of our exploration, the data sources as well as the method of calculation, in some detail and point out the differences to the dominant approaches in the literature.
Selection of countries and elections
The selection of countries and elections constitutes the first important step of our research design. As the implications of globalization, and the degree and the timing of political change, depend on national specifics (see Chapter 2), we have chosen to conduct a comparative analysis of several countries that controls for explanatory factors at the national level. Nevertheless, there are also pragmatic arguments that guide this step of analysis, concerning, among other things, the heavy workload of the content analysis (see below), other available data sets, and last but not least the language skills of the authors and collaborators. In order to analyze the impact of globalization on the national political space in Europe, we finally selected six West European countries: France, Austria, Switzerland, the Netherlands, the United Kingdom and Germany.
We did not include East European countries in our sample because their democratization did not take place until the beginning of the con- temporary era of globalization so that we cannot compare the political mobilizations before and after this crucial development. East European party systems and voter alignments are also rather fluid, which makes comparisons with established liberal democracies difficult. As regards Western Europe, we discussed including Italy in our analysis but we finally decided against it because political change in this country and the breakdown of the First Republic’s party system in the early 1990s is more the consequence of national idiosyncrasies than elsewhere (see,
The design of the study
55
1
e.g.,Newell2000:177–8). OtherSouthEuropeancountries,especially
Spain, democratised only in the 1970s, which makes temporal compari- sons again problematic. Scandinavian countries, finally, are not included in our sample because of lacking language skills of the authors.
As was shown in Chapter 2, the six countries we finally selected are very similar in many respects, expressing a most similar systems design. All six are stable liberal democracies with consolidated political institu- tions and party systems, and all of them belong to the economically most developed and richest countries worldwide. More generally, the set of societal conditions (cleavage structures, economic and cultural context conditions) has created broadly similar latent political poten- tials in all six countries. However, as argued in the previous chapter, they also present some systematic contextual variations, and the political conditions for the mobilization of these potentials vary con- siderably from one country to the other.
After choosing the countries, the second question concerned the kind
of elections to be included. In general, election campaigns provide a
perfect opportunity to study the major lines of conflict in a society,
because the parties, still the dominant political actors in liberal democ-
racies, are forced to express their opinion on all important issues in a
rather short period of time in order to mobilize their electorate and
win new voters. Thus, the important conflicts become visible for the
researcher. Our comparative analysis focuses on national elections
where we explore the changing conflict structures operationalized as
combinations of the issue-positions of voters (the demand side of electoral
competition) and of parties (the supply side) expressed in the context
2
ofelectioncampaigns. Weconsidernationalcontestsstilltobedecisive for the structuring of the political space. They are more appropriate for our research purposes than European elections, as the latter are mostly second-order national elections (Van der Eijk and Franklin 1996). Comparing, for example, turnout in national and European elections, it is clear that the voters regard the former as much more important; the parties too share this perspective as they clearly invest more resources in national campaigns.
1 Additionally, including Italy in our sample would have led to problems regarding available data for the demand analyses.
2 This also means that we concentrate on the national level of party competition. Various sub-national configurations, as for example in East Germany or in the United Kingdom (e.g. in Scotland), are not dealt with.
56 Martin Dolezal
Figure 3.1 The rise of globalization, 1970–2003
Source: KOF Globalization index 2006 (ETH Zurich); see www.globalization- index.org. This index averages the scores of 23 economic, social and political indicators of globalization in 123 countries. Economic indicators include, for example, trade flows, foreign investments and import barriers. Social indicators include tourism, outgoing telephone calls or Internet usage. Political globalization, finally, is operationalized with membership in international organizations and diplomatic relations with foreign countries.
Having decided which countries and what kind of elections we ana- lyze, the last step of the selection process concerned the time period. Globalization is not an entirely new phenomenon, but, because several aspects of the processes related to it accelerated in the late 1980s (see Figure 3.1), we chose to focus primarily on the elections since the 1990s, when the political implications of denationalization became part of partisan politics. In order to analyze the changing conflict structures influenced by globalization processes, we therefore study four elections in each country: three from the 1990s and early 2000s and one from the mid-1970s. The latter serves as a point of reference from a period before the national politics were undergoing the presumed restructuring effect of globalization.
We include several elections from the 1990s in our analysis because we assume, in line with a renewed realignment-theory (Martin 2000), that a structural transformation of national politics may occur across a
The design of the study
57
Table 3.1 Elections
Point of reference in the 1970s
Elections in the 1990s/2000s
1988 1995 2002 1994 1999 2002 1991 1995 1999
Francea Austriab Switzerland Netherlands UK Germany
1978
1975
1975
1973
1974 (February) 1976
1994 1998
1992 1997 2001 1994 1998 2002
2002 and 2003
a 1978 election to the parliament; 1988, 1995 and 2002 presidential elections
(as explained in the text).
b We do not analyze the snap election of 1995 because this campaign was dominated to an extraordinary extent by the major parties.
series of critical elections over an extended period of time. In all countries but France, we analyze the national parliamentary elections. In France, the presidential elections are more important (Bell 2000: 1; Knapp 2004: 25), which is why we selected them for our exploration, with the excep- tion of the election in the 1970s. Because no national election study was available for the presidential races in the 1970s, we exceptionally chose the parliamentary elections of 1978 instead. Table 3.1 lists all elections selected for this study.
Two sides of party competition: data collection and analysis
Throughout this book we distinguish between two sides of party competition: the demand side, covering the political preferences of the electorate, and the supply side, summarizing the issue-positions of the parties competing in the elections. Both sides determine the changing conflict structures in European societies. In order to analyse these structures, i.e. the configurations of the issue-positions of voters and parties in the national political space, we constructed a system of twelve issue-categories that cover (almost) all political subjects in Western democracies. These categories will be used for the analysis of both the demand and the supply side. After the introduction of these categories, detailed information will be given on how we collected and analyzed the data.
58 Martin Dolezal Twelve issue categories
Analyzing the issue-attitudes of voters and issue-positions of parties builds the core of our study. For the supply side, we coded the issue- positions with great detail (see below), but, for the analysis, we had to regroup the sometimes more than 200 codes into more encompassing categories. This step is important for both theoretical and technical reasons. From a theoretical perspective, the specific issues raised during a campaign vary from one election to the next as a result of the policy attention cycle, which in turn depends on the development of the policy- making process in the various political subsystems of a given polity (see van der Brug 1999, 2001). Issues may come up on the electoral agenda as a result of internal dynamics in certain political subsystems or as a result of external shocks, catastrophes (such as September 11 in 2001, the flood in Eastern Germany in 2002 or the war in Kosovo in 1999) or economic crises. Although the specific issues raised during a given campaign are, therefore, somewhat unpredictable, they still refer to only a limited set of basic structural conflicts, which they articulate in various ways. The theoretical challenge is to regroup these issues into a limited, but exhaustive, set of basic categories that is capable of captur- ing the underlying dimensions of a conflict and generally applicable for the analysis of a longer time period. Not all national particularities could be taken into consideration. In Germany, for example, we had coded several issues connected to the country’s reunification, but, for the comparative analyses, we did not retain a specific code for reunification. Technically, we also need a limited set of categories in order to have enough cases per category for all elections covered (see below).
In both parts of the analysis, the regrouping of individual, some- times very detailed, issues into the twelve categories is therefore per- haps the most difficult part of the exploration because these decisions heavily influence all further steps of the analysis. As we are interested in how parties and voters respectively position themselves with regard to political issues (direction), it is important to distinguish between support and opposition; all categories are therefore defined in such a way that they include a direction. Table 3.2 lists and describes these twelve categories and includes the abbreviations that will be used in the figures. This system of categories will also be used when analyzing the preferences of voters (the demand side) to enable the comparison of supply and demand, which is a crucial point of our analysis.
The design of the study
59
Table 3.2 Issue-categories Abbreviation in
Category Welfare
Budget
Economic liberalism
Cultural liberalism
the figures welfare
budget
ecolib
cultlib
Description
Support for an expansion of the welfare state; defence against welfare state retrenchment; support for tax reforms with a redistributive character; calls for employment and health care programmes
Support for a rigid budgetary policy; reduction of the state deficit; cuts in expenditures; reduction of taxes without direct effects on redistribution
Support for deregulation, more competition, and privatization; opposition to market regulation; opposition to economic protectionism in agriculture and other sectors of the economy
Support for the goals of new social movements, with the exception of
the environmental movement; support for cultural diversity, international cooperation (excluding the European Union and NATO); support for the United Nations; opposition to racism; support for the right to abortion and euthanasia; opposition to patriotism, calls for national solidarity, defence of tradition, national sovereignty, and to traditional moral values; support for a liberal drug policy
Support for European integration, including enlargement; support for EU membership in the cases of Switzerland and Austria
Support for education, culture, and scientific research
Support for a tough immigration and integration policy
Support for the armed forces (including NATO), for a strong national defence, and for nuclear weapons
Europe
Culture Immigration Army
europe
culture immigration army
60
Martin Dolezal
Table 3.2 (cont.) Abbreviation in
Category Security
Environment
Institutional reform
Infrastructure
the figures security
environ iref
infra
Description
Support for more law-and-order, the fight against crime, and denouncing political corruption
Support for environmental protection; opposition to nuclear energy
Support for various institutional reforms such as the extension of direct democracy, modifications in the structure of the political system, federalism and decentralization; calls for the efficiency of government and public administration, and new public management
Calls for the improvement of the country’s roads, railways, etc.
The first three categories – welfare, budget, and economic liberalism – refer to the traditional economic opposition between state and market, which is the class-based opposition between left and right. The next six categories – cultural liberalism, Europe, culture, immigration, army, and security – all refer to the cultural dimension of societal conflicts. Finally, there are three additional categories that do not ‘automatically’ belong to one of the two major dimensions of conflict: environment, institutional reform, and infrastructure. However, whether the twelve issue categories indeed build two basic lines of conflict, and which categories belong together, are always empirical questions.
Analysis of the demand side
For the exploration of the demand side, we rely on secondary analyses
of existing surveys. In all six countries but Austria, national election
studies are available and provide detailed information concerning
most of our research questions. In the Austrian case, we depend
on international surveys for the elections of 1975 and 2002, and on
3 3 Also, the Swiss survey of 1975 was part of an international study (see Table 3.3).
less perfect national polls for the elections of 1994 and 1999.
The design of the study 61
Table 3.3 lists all surveys used for this study; detailed information on the variables is given in Appendix A.
Two basic questions guide the exploration of the demand side of party competition: on the one hand, we want to analyze the structure of the political space based on the voters’ attitudes; on the other hand, we are interested in the positions of various social groups in this space: voters of the different parties, members of social classes, groups defined by their level of education and religious affiliations. As discussed in the introductory chapters, according to our theory, social class as well as the level of education are the most important features distinguishing between winners and losers of globalization.
To make the analysis of the demand side comparable between elections
and countries as well as with the analysis of the supply side, we use the
same twelve issue-categories introduced above for both sides of the study.
In a first step, in each survey all variables measuring issue-positions4 of
the voters (or issue-priorities if no positions were asked for) were iden-
5
variable was available for a given category, we factor-analyzed them and
used the common factor instead. In some cases, however, the issues we
associated with a given category on theoretical grounds did not all load
on a single common factor; in these situations, a second category had to
be constructed. These exceptions will be explained in the country chap-
ters. The variables included in the different surveys typically did not allow
covering all categories, i.e. we usually did not have information on every
aspect we are interested in. In particular, attitudes about the budget and
infrastructural projects were very rarely asked. The most important
categories are, however, covered by most surveys. Table 3.4 summarizes
the issue-categories we are able to measure on the basis of the available
data. In a final step, we performed factor analyses (with varimax rota-
tion) of the available categories that always resulted in two-dimensional
6
4 Because we are interested in issue-positions, we did not select left–right scales, materialism–post-materialism scales, and all variables that seemed too vague, such as questions regarding the importance of ‘solidarity’ or the like.
5 For details, see Appendix A.
6 For exceptions, see the country chapters.
tified and regrouped into our system of categories. If more than one
solutions.
calculation constitute the basis for the graphical presentations concerning the demand side in the following chapters, which show the spatial posi- tions of party voters and those social groups we are especially interested
These configurations and the factor scores derived from the
Table 3.3 Surveys used for the demand-side analyses
France
Austria
Switzerland
1978: Enquête post-électorale française, 1978 (q0062) 1988: Enquête post-électorale française, 1988 (q0601)
1975: Political Action – An Eight Nation Study
1975: Attitudes politiques 1975 (20) (part of ‘Political Action – An Eight Nation Study’)
1995: Enquête post-électorale française, 1995 (q0891)
1994: Sozialwissenschaftliche Studiengesellschaft 9409 1999: Sozialwissenschaftliche Studiengesellschaft 9908
1991: not availablea
1995: Swiss electoral study 1995 (1815) 1999: Swiss electoral study 1999 (6646)
2002: Panel électoral français 2002 (PEF 2002)
2002: European Social Survey 2002–2003 UK
The Netherlands
Germany
1972: Dutch parliamentary election study, 1972, 1973 (P0353)
1974: British Election Study, February 1974 (UKDA study number 359)
1976: Wahlstudie 1976 (ZA0823) 1994: Nachwahlstudie 1994 (ZA 2601) 1998: Politische Einstellungen, politische
1994: Dutch parliamentary election study, 1994 (P1208)
1992: British General Election Study, 1992 (UKDA study number 2981)
Partizipation und Wählerverhalten im
1998: Dutch parliamentary election study, 1998 (P1415)
1997: British General Election Study, 1997 (UKDA study number 3887)
vereinigten Deutschland 1998 (ZA 3066) 2002: Bundestagswahlstudie 2002 (ZA 3861)
2002–3:Dutchparliamentaryelection study, 2002–2003
2001:BritishElectionPanelStudy,2001 (UKDA study number 4620)
a For 1991, there are insufficient issue-questions in the available survey, so we had to exclude this election from the demand analyses.
Table 3.4 Issue-categories covered by the surveys
France Austria Switzerland The Netherlands UK Germany
1978 1988 1995 2002 1975 1994 1999 2002 1975 1995 1999 1972 1994 1998 2002 1974 1992 1997 2001 1976 1994 1998 2002
Welfare xx xxxxxxx x xxxxxx
Budget x
Economic x x x x
x x
x
x x x x x x x x x x x x x x
liberalism
Cultural x x x x x
x x x x x x x x x x x x x x x Europe xx xxxxxxxxxxxxx xx
liberalism
Culturexx
Immigration xxxxxxxxxxxxxx Army x x xxxxxx Security xxxxxxxxxxx xxx
xxxx
xx xxx
Environment Institutional reform
xxxxxxx xxx x x
xx x xxxxx xxxxxxx xx
Infrastructure
x
64 Martin Dolezal
in. To facilitate the interpretation of the corresponding figures, we always rotated the configurations in such a way that the economic left–right dimension is arranged horizontally and the cultural dimension vertically. Pro-state positions are always situated on the left and pro-market posi- tions on the right of the horizontal axis. With respect to the cultural line of conflict, attitudes in favour of integration are always placed at the top, and those favouring demarcation at the bottom of the vertical axis.
The figures for the demand side first include the positions of the voters of the various parties and, when they are not in a centrist position, of non- voters. In general, parties (and all other groups) with less than ten voters in the survey were excluded because the measure of their position was considered unreliable. Apart from the party voters, the figures include the positions of various groups defined by their social class, their level of education, and their religious affiliation. The operationalization of the social classes differs between the explored countries and elections because the available information in the surveys varies widely. Whenever possi- ble, we use a modified version of the Erikson/Goldthorpe class scheme, based on the work of Kriesi (1989, 1998) and Müller (1998, 1999). This classification distinguishes between eight groups: ‘farmers’, ‘other self- employed in non-professional occupations’, ‘semiskilled and unskilled workers including agricultural workers’, ‘skilled workers and foremen’, ‘routine non-manual white collar employees’, ‘managers and other professionals in social-administrative occupations’, ‘professionals with technical expertise’, and ‘social-cultural specialists’. Additionally, we explore the attitudes of voters belonging to the non-labour force. With regard to the level of education, we always compare three groups: ‘low’ education comprises all respondents who completed only compulsory school or less; ‘medium’ education ranges from basic professional train- ing (for example, ‘Lehre’ in the German-speaking countries) to several kinds of high-school graduates (for example, ‘Matura’ in Austria and Switzerland, ‘Abitur’ in Germany); ‘high’ education summarizes all kinds of tertiary education. Finally, concerning the religious cleavage, we com- pare the issue-preferences of Catholics, Protestants and the non-affiliated, as well as the differences between religious and non-religious voters – depending on the country-specific nature of the religious cleavage. A respondent’s religiosity is always operationalized with church-going frequency (see Jagodzinski and Dobbelaere 1995: 87–91).
As explained in Chapter 2, one of our basic hypotheses refers to a ‘zero-sum’ relationship between the strength of old and the potential for
The design of the study 65
new cleavages. In addition to the figures that show the political space, all country chapters therefore include line diagrams indicating the strength of traditional cleavages as well as of new divisions, measured as the difference in the attitudes of their core groups.
Analysis of the supply side
For the analysis of the supply side of electoral competition, we assume that the structural change linked to globalization is articulated by the issue-specific positions taken by the parties in electoral campaigns as well as by the salience they attribute to the different issues. We also consider that the most appropriate way to analyze the positioning of parties and the way in which they deal with the new issues linked with globalization is to focus on the political debate during campaigns, as reflected by the mass media. Our research strategy, therefore, differs from the two so far dominant approaches in the literature: expert surveys (e.g. Castles and Mair 1984; Laver and Hunt 1992; Huber and Inglehart 1995; Benoit and Laver 2006) and the analysis of party manifestos, especially the comparative manifestos project (see Budge et al. 1987, 2001; Klingemann et al. 1994; Volkens 2001).
Expert surveys are a useful and relatively feasible method for estimat-
ing policy positions. But we think there are several major disadvantages
concerning the method per se as well as in relation to our research
question. In general, it remains an open question on which information
experts rely when estimating the parties’ positions (Mair 2001: 24–5).
Sometimes it is also not clear which ‘party’ the experts are positioning:
the party leaders, the activists, the voters, or all three together; in
addition, the designated time period is not always clear (Budge 2000:
103–4). All three objections can be rebutted with a perfect research
design (Steenbergen, Hooghe and Marks 2004), but two additional
factors complicate the estimates: First, the positions of several parties –
especially of small parties – are often difficult to judge because their
attitude concerning issues that are not important for them are often
7
7 Consider, for example, economic or environmental positions of right-wing parties in Germany. Even experts in the field might have difficulties when asked to distinguish the differences between the Republikaner, the DVU and the NPD.
unknown – even to the experts. Secondly, experts tend to deduce a stronger polarization from a higher fragmentation of the party system
66 Martin Dolezal
and therefore provide more widespread estimates for countries with
many parties (Mair 2001: 20–3). With regard to our research question,
finally, developments over time cannot be analyzed with available
data sets based on this method because the survey conducted by Laver
and Hunt (1992) was the first that covered several issues-positions of
the parties and there are no data available on the parties’ positions thirty
years ago. In addition, it is also not possible to operationalize our
system of issue-categories with the available expert surveys, since they
8
tions of parties so far has been the analysis of their manifestos, especially
9
asked too few questions about policy positions.
The most important approach for estimating the programmatic posi-
with human coding.
project (CMP) is the dominant actor in this field of research. The great contribution of this project to comparative politics is the analysis of a large number of countries throughout the post-war period; with respect to the USA, the data set even goes back to the 1920s (Volkens 2001: 34–5). But at least four major problems make it necessary to think of alternatives for getting better information on parties’ positions. First, the CMP is based on the salience theory of party competition, which, in general, neglects the direction of parties’ statements. Extensive research based on party manifestos has, indeed, shown that parties tend to avoid direct confrontation and that they differ from each other mainly through the selective emphasis of their priorities. But we also know that new issues usually do not have a valence character and that direct confrontation – i.e. parties advocating diverging positions on political issues – is much more pronounced in the media and during electoral campaigns than in party programmes (Budge and Farlie 1983b: 281). The voters, too, see the parties mainly in confrontational terms, especially during election campaigns. Secondly, the CMP’s system of issue-categories does not cover important new issues of the 1990s, especially immigration. Statements towards immigration are coded as part of ‘per705 underprivileged minority groups’, and – because of the theoretical base of the CMP – only favourable statements can be included. Thirdly, the overall relevance of party (or electoral)
Without any doubt, the comparative manifestos
8 The choice of policy positions for the survey also strongly influences all further explorations concerning the dimensionality of the space, especially when only few positions are asked for.
9 For computer-generated coding see, e.g., Ray (2001); Garry (2001); and Laver, Benoit and Garry (2003).
The design of the study 67
manifestos is disputable, because voters do not read them. Robertson (1976: 72) rightfully stresses that manifestos provide the basis for statements given by politicians during campaigns, but it is better to directly analyze those statements, which are heard or read by voters; and there is no doubt that contemporary campaigns are primarily fought with mass media. Finally, most manifestos have a kind of soft- focus effect; thus, parties do address several issues but often deliberately avoid clear statements. European integration, for example, is rarely mentioned, and even Eurosceptic parties tend to stress the importance of international cooperation. Furthermore, controversial subjects are sometimes completely excluded from the manifesto. Existing program- matic differences between parties, therefore, tend to be blurred when relying on their manifestos. The mass media, on the contrary, often intensify statements of politicians, which makes it easier to observe the basic political conflicts in a given country.
These methodological deficits of existing approaches led us to the
conclusion that it is necessary to produce a new data set. In order to
identify the salience of the campaign issues for the various parties and
their issue-specific positions, we therefore conducted an extensive content
10
For each electoral campaign, all articles, except commentaries, related to the electoral contest or to politics in general were selected in both newspapers for the last two months before election day. Because political advertisements are very important in Switzerland, we included them in our analysis of this country. With regard to the quality newspapers, we used a sampling method and only selected the articles on three days (for details see the appendix). Because the tabloids have fewer articles, we selected all issues within the two-month period. In general, the sampling procedure was conducted in a way to get the same amount of informa- tion on the parties’ positions in each country. For the articles selected, the
10 The coding was done by some of the authors as well as by students and student research assistants.
analysis of articles in major daily newspapers based on human coding. For each country, we chose a quality newspaper, which can be inter- preted as the leading medium of political coverage, and a tabloid to get a broader picture of how voters see the parties during the campaigns. We always selected the most widespread paper in each category among the newspapers that were published throughout the whole period covered by our study (Table 3.5).
68
Martin Dolezal
Table 3.5 Selected newspapers
Quality paper
Tabloid
Le Parisien
Kronenzeitung
Blick
a
The Sun Bild
France
Austria Switzerland
The Netherlands
UK Germany
Le Monde
Die Presse
Neue Zürcher Zeitung NRC Handelsblad
Algemeen Dagblad The Times Süddeutsche Zeitung
!
a In the Netherlands, no widespread tabloid exists.
headlines, the ‘lead’, if available, and the first paragraph11 were coded sentence by sentence using a relational method of content analysis devel- oped by Kleinnijenhuis and his collaborators (see Kleinnijenhuis et al. 1997; Kleinnijenhuis and Pennings 2001). This method is designed to code every relationship between ‘political objects’ (i.e. either between two political actors or between a political actor and a political issue) appearing in the text. For the purposes of this study, we are only interested in relationships between political actors and political issues (‘actor-issue sentences’). According to this coding procedure, each sentence of an article is reduced to its most basic structure, the so- called ‘core sentence’, indicating only its subject (political actor) and its object (issue) as well as the direction of the relationship between the two. The direction is quantified using a scale ranging from !1 to +1 with three intermediary positions indicating a ‘potential’ or a neutral relation. If, for example, a politician says that in the future he might be in favour of or against a certain position we coded 0.5 or !0.5 respectively. Whenever there was a neutral relation – that is no direc- tion at all – we coded 0. The issues were recorded with great detail; later we aggregated them to our system of issue-categories as described above. Political actors were coded according to their party membership. For the analyses in the subsequent chapters, we have regrouped them into a limited number of categories or observed just the most important parties respectively, from three in the UK to eight in France (Table 3.6).
11 For some tabloids we coded the complete articles because they are very short.
Table 3.6 Parties according to current membership of party families
France
Parti Communiste Français – PCF (French Communist Party) and Trotskyist parties
Les Verts (The Greens)
Parti Socialiste – PSF (Socialist Party) Mouvement des Radicaux de Gauche – MRGa (Movement of Left Radicals) Sozialdemokratische Partei
Union pour la Démocratie Française – UDFb (Union for French Democracy)
Rassemblement pour la République – RPR (Rally for the Republic)
Austria
Die Grünen/Die Grüne
Liberales Forum – LIF (Liberal Forum)
Österreichische Volkspartei – ÖVP (Austrian People’s Party)
Freiheitliche Partei
Switzerland
Grüne Partei der Schweiz – GPS (Green Party of Switzerland)
Sozialdemokratische Partei der
Schweiz – SPS (Social Democratic Party of Switzerland)
Freisinnig- Demokratische Partei – FDP (Radical Democratic Party)
Christlichdemokratische Volkspartei der Schweiz – CVPd (Christian Democratic People’s Party)
Schweizerische Volkspartei – SVPe (Swiss People’s Party)
Communists/ Radical Left
Greens/ New Left
Social Democrats
Liberals
Christian Democrats/ Conservatives
Populist Right Front National
Alternative (The Greens/ The Green Alternative)
Österreichs – SPÖ (Social Democratic Party of Austria)
Österreichs – FPÖc (Freedom Party of Austria)
Liberale Partei der Schweiz – LPS (Swiss Liberal Party)
– FN (National Front)
Table 3.6 (cont.) Communists/
Greens/ New Left
Social
Democrats Liberals
Christian Democrats/ Conservatives
Netherlands
GroenLinks (Green Left)
PartijvandeArbeid– Democraten’66–
Christen-Democratisch Appel – CDA (Christian Democratic Appeal)
UKf Germany
Labour Party
Liberal Democratic Party
Conservative Party
Radical Left
Populist Right Lijst Pim
Partei des demokratischen Sozialismus – PDS (Party of Democratic Socialism)
Bündnis 90/Die Grünen
Sozialdemokratische Partei
Freie Demokratische Partei – FDP (Free Democratic Party)
Christlich-Demokratische Union/Christlich-Soziale Union – CDU/CSU (Christian Democratic Union/Christian Social Union)
(Alliance 90/ The Greens)
Deutschlands – SPD (Social Democratic Party of Germany)
PvdA (Labour Party)
D’66 (Democrats
Fortuyn – LPF (List Pim Fortuyn)
a It is difficult to classify the MRG into one of the major party families. It is rather left-of-centre and was one of the three ‘pillars’ of the left-wing opposition in the 1970s (with the PCF and the PSF). But it cannot simply be subsumed into the social-democratic party family.
b The UDF has both a liberal and a Christian-democratic component.
c Until the mid-1980s, the FPÖ was a liberal-conservative party.
d Plus other minor centre parties.
e Plus several small parties of the radical right. Until the early 1990s, the SVP was part of the conservative party family. f The Scottish National Party (SNP), which is difficult to classify in this table, is also included in the analyses.
66) Volkspartij voor
Vrijheid en Democratie – VVD (People’s Party for Freedom and Democracy)
The design of the study 71
We cannot, however, consider all parties in all elections. As a matter of fact, some of them did not exist during the whole period. The Lijst Pim Fortuyn, for example, was only present in the Dutch elections of 2002 and 2003; the Liberal Forum, a splinter party of the FPÖ, was visible only in the Austrian elections from 1994 to 1999, then it became irrelevant – for voters as well as for the media. Some smaller parties also had to be excluded, as we do not have enough information on their issue positions. In each election, we mostly consider only those parties for which at least thirty issue positions were coded. Especially in Germany, parties of the radical right are not given coverage in newspapers, at least there is no information about their issue-positions (apart from immigration). Such parties, therefore, could not be included in our supply side analyses, which is one of the disadvantages of our methodological choice. But we are able to explore their voters’ attitudes in the demand analyses.
The data resulting from our content analysis offer valuable informa- tion on two central aspects of the supply side of electoral competition: the positions of political parties regarding the various political issues, and the salience of these issues for a given political party. The position of an actor on a category of issues is computed by averaging over all core sentences that contain a relationship between this actor and any of the issues belonging to this category. The salience of a category of issues is based on the relative frequency with which a given political party takes a position on this category. As argued above, it is important to understand that both aspects are relevant for an adequate description of the political space. Parties differ from one another not only with respect to the posi- tions they advocate, but also with respect to the priorities they set. It is also important to note that the salience of issues and parties can be computed in different ways. In this study, party-issue relationships are weighted by the number of statements of a given party in a given cam- paign and by the relative importance of the corresponding issue category for the party in question. This means that, for a given campaign, large parties and key campaign issues determine the configuration of the political space more heavily than marginal parties or secondary issues.
On the basis of these data, it is finally possible to construct a graphical representation of the positions of parties and issues in a low-dimensional space, using the method of multidimensional scaling (MDS). MDS is a very flexible method that allows for the graphic representation of similarities or dissimilarities between pairs of objects (see, e.g., Borg and Groenen 1997; Cox and Cox 2001; Kruskal and Wish 1978). If a
72 Martin Dolezal
party from the left, for example, strongly supports an expansion of the
welfare state, we would expect the distance between this party and the
category welfare to be small. If we represent the parties and issues in a
common space, this party and the category welfare should be located
close to each other. The unfolding technique, the MDS procedure which
we use here, allows for the joint representation of parties and issues in a
12
common space.
Furthermore, a variant of MDS, called weighted
metric multidimensional scaling (WMMDS), enables us to account
simultaneously for the similarities between pairs of objects (parties
13
and issues, in our case) and for the salience of these relationships. This means that, when representing our data in a low-dimensional space, the distances corresponding to salient relationships between parties and issues will be more accurate than the less salient ones. Distortions of ‘real’ distances are unavoidable, but with WMMDS these distortions will be smaller for more important relationships, resulting in a more accurate representation of the relative positions of parties and issues. Relying on MDS has an additional advantage crucial for our argument. With this method, we do not have to make any a priori assumption about the structure of the political space – contrary to most analyses of parties’ positions that start from theoretically defined dimensions (see, e.g., Gabel and Hix 2002; Hix 1999; Hooghe et al. 2002; Klingemann et al. 1994; Pellikaan et al. 2003; Pennings and Keman 2003; van der Eijk and Franklin 2004). Here, by contrast, and similar to our strategy for the demand side analyses, we want to test our hypotheses regarding both the dimensionality of the political space and the nature of these dimensions. The structure of the political space that we estimate with WMMDS is not influenced by any assumption we could make on how the categories of issues are related to one another.
When interpreting the supply-side configurations and comparing them with the demand-side figures in the following chapters, it is important to take some technical differences into consideration. Just as the demand-side figures, the supply-side figures are rotated so that
12 Van der Brug (1999, 2001) uses another MDS procedure that does not allow for the joint representation of parties and issues in a common space, which makes the interpretation of the results much more difficult.
13 Weighted metric multidimensional scaling can be estimated using the algorithm Proxscal, which is implemented in SPSS.
The design of the study 73
the economic conflict is arranged horizontally and the cultural conflict vertically; again, positions supporting cultural integration are placed on the top, those supporting demarcation at the bottom of the figure. But, in an MDS figure, the crucial aspect refers to the relative distances between points (for example, between a certain party and a certain issue category); their spatial positions are of lesser importance. If, for example, a party is situated on the bottom right side of a configura- tion, it does not necessarily strongly support market liberalism. One has to check its relative distance from economic liberalism and welfare respectively. Only if the party is much closer to economic liberalism can its position be interpreted as support for an economically right policy.
Conclusion
It was the purpose of this chapter to describe the design of our study, the data sources, and the methods applied. Our basic decision was to analyze changing cleavage structures in West European societies based on an observation of issue-positions of parties and voters in the context of electoral campaigns. We explained the comparative approach of this book and argued why we selected the six countries covered by our exploration and why we focus on national elections. Our sample of West European countries – France, Austria, Switzerland, the Netherlands, the UK and Germany – is quite homo- geneous but it provides considerable variation with respect to the national context factors that may explain how the processes asso- ciated with globalization can be mobilized by political entrepreneurs, especially in connection with national elections. Distinguishing between demand and supply as the two sides of party competition, we explained the kind of data we use and how we construct our indicators based on these data. As regards the demand-side analyses, we rely on secondary analyses of available surveys. Concerning the supply side, we have built a new data set based on an extensive relational content analysis of the media coverage of elections. Both parts of the analysis use the same system of issue-categories that provides information on the direction of voters’ political attitudes and the parties’ programmes respectively, as well as on the salience of these issues. Once again, it is important to note that we start our analysis with as few a priori assumptions as possible, because
74 Martin Dolezal
the changing structure of the political space – in both dimensions of competition – is the core question of this book. With the statistical procedures selected – factor analysis for the demand side and weighted multidimensional scaling for the supply side – we use mathematical methods that are perfectly suited for such an explorative research strategy.
