Do all roads lead to the centre? The unresolved dilemma of centrist self-placement Toni Rodon Casarramona Universitat Pompeu Fabra June 14, 2012 Abstract Why do people locate themselves on the centre? In parallel with the debate about which components prevail when citizens locate themselves on the left-right scale, some research has put forward some hypotheses that could explain centrist self-location. However, our knowledge about centrist placement is still surprisingly low, although it is the most numerous position in almost all the countries. This article tests the possible existence of the centrist exceptionalism by testing whether citizens locate themselves on the centre because of the party component, the lack of political sophistication or the salience given to postmaterialist values. The analysis is performed in 52 elections in 21 different countries. It reveals that, above all, the centre position is genuine, which implies that people locate themselves there because it is simply a position between what left and right stand for. Findings in this article have implications for our understanding of the left-rigt axis and its role in party competition. Toni Rodon Casarramona PhD Candidate Universitat Pompeu Fabra c/ramon Trias Fargas, 25-27, 08005 Barcelona, Catalonia Phone: +34 93 542 22 56 Email: toni.rodon@upf.edu Website: www.tonirodon.cat Paper prepared for presentation at the XXII World Congress of Political Science of the International Political Science Association, Madrid, 8-12 July 2012. Number of words: 7,938.
1 Introduction In 1990 the Italian National Elections Survey asked on which ideological position would the respondent locate herself. This question, generally common in almost all the electoral surveys, was followed with another very uncommon question restricted to those that located themselves on the centre of the scale. The wording was the following: You placed yourself in the middle of the row of boxes. Why did you place yourself there?. 50.5% of the Italians answered I strongly support a central position, whereas 49.5% chose the option I don t really identify with the left/right distinction. This very uncommon question wanted to tackle the meaning of the obscurest of the ideological positions, which has puzzled political theorists and methodologists alike. In other words, if it is fairly easy (or less difficult) to define what is to be leftist and what to be rightist, what does to be centrist mean? Definition of left and right has puzzled political scientists for a long time. The strong normative and historical connotations behind both ideological frames has triggered an enormous quantity of research and intellectual debate. However, fewer energies have been devoted to the study of the centre position. Only Knutsen (1998b) put forward three hypotheses that could explain the reason behind the centrist self-placement. In his work, Knutsen suggested that the centre could be a genuine-moderate position or a product of the lack of political sophistication or of the salience given to postmaterialist issues. However, the limitation of the number of countries, the time span of the analysis and the methodological approach adopted in his work left many doubts regarding the real reason of the centrist location. This paper takes insights of Knutsen s pioneering work and tries to fill the gap by tackling the following question: Why do people locate themselves on the centre position?. By answering this question, I contribute to the debate over ideological self-placement by testing again Knutsen s hypotheses, which are, furthermore, combined with other hypotheses that the ideological self-placement literature put forward a long time ago. In other words, I integrate Knutsen s insights and other hypotheses that can potentially explain centrist self-placement. The empirical analysis takes a broader perspective than previous works, contemplating 52 country-elections and thus several contextual scenarios. Findings in this article show that in several cases centrist self-placement emerges as a product of party proximity or to the lack of political sophistication. However, the impact of both factors is generally weak. On the other hand, the salience given to 1
postmodern values do not increase the likelihood to be on the centre of the scale. Overall, the hypothesis that seems to have more credit is the centre as a genuine position, that is, an ideological position between left and right. 2 Ideological self-placement Since the French Revolution coined the terms in 1789, the language of left and right has been one of the -if not the- most popular concepts in any political action. From sociology to philosophy, very few concepts can claim to overarch the field, both academically or mediatically. In modern politics, the left-right divide has served as a political schema classifying ideologies that has allowed parties to communicate with voters and the latter to orient themselves in a complex political world (Converse, 1964). It is perceived as a super-issue that reduces differences in party positions on a broad range of issues to be at least minimally compatible and consistent with each other (Castles and Mair, 1984; Hamilton, 1987; Huber and Inglehart, 1995; Hinich and Munger, 1996; Laver and Schofield, 1998; Knight, 2006; Budge, Robertson and Hearl, 2008; Gschwend and Proksch, 2010). As such, it helps voters economise on information and to choose a party to represent them (Downs, 1957; Lau and Redlawsk, 2001; Blais and Nadeau, 2004; Franklin, Mackie and Valen, 2009). One of the most intensive debate around the ideological concepts, still unresolved, has focused on what the true meaning of left and right is. Since both dimensions gained popularity and started to be employed as political tools, a very fruitful and ongoing research has dedicated its efforts to identify what left and right mean for the general public. In this sense, the number of answers is probably as high as the number of articles published dealing with this issue. Downs (1957) influential work probably triggered the modern debate when he focused on the level of government intervention in the economy, which was a concept later aligned with the traditional left and right dimension. A few years earlier, however, Lipset and Linz (1954, 1135) already set a precedent defining the ideological axis in terms of equality (left) and hierarchical social order (right). Since then, left and right have been defined in terms of values (Inglehart and Klingemann, 1976); equality versus liberty (Van Deth and Geurts, 1989; Bobbio, 1997; Jahn, 2011); new political issues such as the environment or citizens political participation (Powell, 2000); as a tension between cultural aspects and different meanings of egalitarianism (Elchardus and Spruyt, 2012); or even the 2
alignment with certain valence issues (Zechmeister, 2006). These are only some examples that exemplify the never-ending debate over left and right definition. This somehow philosophical debate over the meaning of left and right has not been the unique approach in order to know what both concepts stand for. Since the influential work of Inglehart and Klingemann (1976), a different approach has been adopted: if we are able to answer why citizens locate themselves on the left-right axis, we will be capable of knowing the meaning attach to them. Although similar in nature, this approach is different because, instead of identifying the true philosophical or historical meaning of left and right, it tries to distinguish the individual microfoundations by which citizens locate themselves on a certain position. Following this reasoning, the literature identifies three components that can explain individuals ideological position: first, the social component. Second, the attitudes (or issue) component. Third, the partisan component. The social component refers to individuals location in a social surrounding which corresponds with their social identity and their left-right opinion. Inglehart and Klingemann (1976, 264-269) already argued that the impact of the social component is rather small, especially when compared to the partisan dimension. Even though some differences seem to exist between older and younger generations regarding the effect of this component (Freire, 2006), the fact is that the attitudes of an individual (second component), as well as his partisan preferences (third component), are largely dependent on the social position of this individual (Weber, 2011). Therefore, by including the following two components, the social component is also contemplated in an indirect way. The second component, the value or issue component, departs from the idea that the left and the right are ideological concepts with strong normative connotations. Therefore, when deciding which position the individual wants to locate himself on, 3
values and attitudes emerge as a crucial driving factor 1. As sketched before, the proponents of the value component depart from different preconceptions of what left and right can mean and try to see whether individuals take into account these conceptions when locating themselves on the scale. Finally, the partisan component refers to the existence of a party of reference that drive individuals to locate themselves on a particular position. This theory stresses that partisanship is the primary component of left-right self-placement. It is well established that individuals have low political information and they need cues in order to interpret and analyse political events (Lau and Redlawsk, 2001). In this sense, political parties provide useful tools for individuals to develop and defend opinions and they can act as an important heuristic actor for ideology. Therefore, people s position on the left-right scale will not be a reflection of their social class or their values or attitudes. It will rather be a consequence of parties ideological distribution on the left-right scale (Adams, Merrill and Grofman, 2005). For instance, people will decide to locate on the centre of the scale if -and only if- they have a party of reference which is also located on the centre or close to this position. Recent literature has shown that this component is generally the one that weighs more when deciding the individual s ideological position (Huber, 1989; Knutsen, 1997, 1998c; Zechmeister, 2006; Medina, 2010). These findings do not imply, however, that the debate is closed. The most important shortcoming in the articles dealing with the partisan component is the assumption that parties matter, regardless of their ideological position. Generally, the study of the partisan component studies those that locate on the scale versus those that do not locate on the scale. This simplification goes actually against the idea of the partisan component, as party distribution along the axis may have a different impact along the ideological scale. For instance, in a given country where no extreme left-wing parties exist, the partisan component will certainly not matter for those located on this position, as they do not have any party of reference. As I explain in the next subsection, this assumption 1 To be fair with the literature, some studies consider that values and attitudes to issues are two different concepts (Weber, 2011). The idea is that values and attitudes are long-term factors and, in a causal sense, they are prior to issue preferences. Values are considered relatively stable, whereas attitudes to issues are less stable and can change when different issues are salient in the public sphere. This literature argues that not all issues serve every time as indicators for values and that issue preferences may be ad hoc reactions to the current political debate but do not actually reflect an underlying value. Even though this idea should be considered with more sympathy (Milic, 2008), in this article both concepts are used alternating. The main reason is that it is not the goal of this paper to establish a causal connection -both in direction or intensity- between values and issues. At the same time, with the type of data used in this article it is not possible to disentangle both concepts, that is, to find issues without any value connotation. 4
emerges as crucial when dealing with the centre position. To sum up, according to the ideas explored above, people locate themselves on the scale, first, because of his/her attitudes or values and, second, because the party operates as an ideological mirror and the individuals locate themselves on the scale according to his/her preferred party position. 2.1 The centrist puzzle Note that in the previous subsection things have been simplified and one important concept has been ignored: the centre. It would not be a problem if what Duverger (1964, 215) stressed was true: In politics, the centre does not exist. However, Duverger s assertion has proven to be far beyond what is common in any political struggle. Despite its popularity in the media, the concept of the centre has remained generally unexplored in the literature. This curious anomaly brought Daalder (1984, 92) to highlight that there is practically no systematic treatment of the centre or of center parties 2. This fact does not only stem from the difficulty to decide which approach is most appropriate to study the centre (Duverger, 1964; Daalder, 1984; Sartori, 2008), but to the idea that identifying what the centre is becomes much more difficult than grasping the true meaning of the left and the right. Almost thirty years after Daalder s work, the situation remains very similar. Scholars tend to think of left and right as dichotomous concepts with no point in-between 3. However, for more than fifty years, left-right scales -the empirical operationalisation of the left-right dimension-, which have a more or less a clear mid-point, have been widely used in surveys (Inglehart and Klingemann, 1976; Kroh, 2007). Not only a centrist point is usually offered in surveys, but, when taken alone, it 2 There are some exceptions (Daalder, 1984; Hazan, 1998; Keman, 1994), although they focus their attention to the notion of centrist parties rather to the ideological centre. 3 The 1994 Italian National Study provides a clear example of this tendency to dichotomise the political ideology. In this survey, after the vote intention question was asked, the following question was included: If the party mentioned is on the centre: is it towards right or towards left?. 60.62% of the respondents answered: Centre-centre. Later, the survey asked again the same question but in this occasion it referred to the individuals self-location on the centre. The answer centre-centre was again the most numerous (64.14%). Certainly, these questions would not have been asked (and this is probably why it were not included), had the answer referred to a leftist or rightist party or position. That is, for some scholars a centrist answer may trigger the feeling that the respondent does not want to answer his/her true ideology. 5
emerges as the most populated position in almost all European countries (Medina, 2010). That is, in most of the countries most of the individuals locate themselves on the centre position 4. The following table shows the ideological distribution in different elections in Europe. For the sake of simplicity, the leftist categories (0-10) and the rightist categories (6-10) have been grouped together. Despite this grouping, the centre position (5) is still relevant and in some cases it is the most quantitatively numerous in comparison with all the other ideological positions. Table 1: Ideological distribution in different elections in Europe (%) Country-election Left Centre Right Country-election Left Centre Right Switzerland 1999 30.14 31.31 38.54 Ireland 2002 18.16 36.91 44.93 Czech Republic 1996 26.03 24.60 49.37 Iceland 2003 29.30 28.53 42.17 Germany 1998 57.78 23.21 19.01 Italy 2006 35.86 23.46 40.68 Denmark 2001 28.62 22.98 48.40 Netherlands 2002 36.93 19.60 43.47 Spain 1996 48.99 28.35 22.66 Norway 2001 29.55 22.58 47.87 Spain 2000 37.48 34.27 28.25 Poland 2001 42.49 29.77 27.75 UK 1997 32.87 33.84 33.29 Portugal 2002 30.13 32.46 37.41 Hungary 1998 35.11 33.84 31.05 Portugal 2005 33.53 31.66 34.81 Iceland 1999 25.86 31.04 43.10 Romania 2004 22.48 25.32 52.21 Netherlands 1998 34.97 15.98 49.05 Slovenia 2004 29.35 40.13 30.51 Norway 1997 34.18 25.77 40.05 Sweden 2002 42.44 19.74 37.82 Poland 1997 27.23 31.53 41.24 Austria 2008 37.40 30.71 31.89 Portugal 2002 30.13 32.46 37.41 Switzerland 2007 28.66 31.21 40.13 Romania 1996 17.20 32.12 50.67 Czech Republic 2006 33.53 21.70 44.95 Slovenia 1996 26.41 50.00 23.59 Germany 2005 53.76 20.00 26.24 Sweden 1998 40.06 19.67 40.27 Germany 2009 49.37 24.63 26.00 Bulgaria 2001 17.80 33.46 48.74 Finland 2007 25.08 23.99 50.92 Switzerland 2003 31.66 31.95 36.39 France 2007 30.07 23.48 46.45 Czech Republic 2002 45.57 20.60 33.83 Ireland 2007 21.98 33.08 44.95 Germany 2002 50.69 22.59 26.72 Iceland 2007 27.57 26.20 46.23 Denmark 2001 30.04 22.86 47.10 Iceland 2009 33.96 28.09 37.96 Spain 2004 52.77 25.07 22.16 Norway 2005 36.75 21.31 41.94 Finland 2003 26.64 25.35 48.01 Poland 2005 18.49 31.83 49.68 France 2002 35.13 24.56 40.31 Poland 2007 14.78 31.75 53.46 UK 2005 39.23 33.39 37.38 Portugal 2009 33.88 31.96 34.17 Hungary 2002 41.38 21.93 36.70 Sweden 2006 33.97 18.48 47.55 Comparative Study of Electoral Systems (CSES). Module 1, 2 and 3. The following graph shows the percentage of people self-located on the centre ordered by country-elections. 1996 and 2004 Slovenia elections are the cases where most people located themselves on position 5, i.e. the centre of the 0-10 ideological scale (50% and 40.13%, respectively). On the other extreme, the 1998 Dutch elections and the 2006 Swedish elections where the ones with fewer people self-located on the centre of the scale (15.98% and 18.48%, respectively). Again, these graphs stress the idea that, despite the variation of the percentage of people located on the centre across 4 Even when the scale employed in some surveys does not have a centrist point, a centrist tendency is also reported. In this case, number 5 still seems to operate as a centrist category, although in the 1-10 scales such a position does not exist because of the uneven numbers of the scale. 6
countries, this position is, with some exceptions, the most populated ideological position, which turns it as the a priori most coveted position by political parties (Downs, 1957). Figure 1: Centrist self-placement in different elections across Europe Slovenia 1996 Slovenia 2004 Ireland 2002 Spain 2000 Hungary 1998 UK 1997 Bulgaria 2001 UK 2005 Ireland 2007 Portugal 2002 Portugal 2002 Romania 1996 Portugal 2009 Switzerland 2003 Poland 2005 Poland 2007 Portugal 2005 Poland1997 Switzerland 1999 Switzerland 2007 Iceland 1999 Austria 2008 Poland 2001 Iceland 2003 Spain 1996 Iceland 2009 0 10 20 30 40 50 Individuals self located on the centre (%) (a) Norway 1997 Finland 2003 Romania 2004 Spain 2004 Germany 2009 Czech R 1996 France 2002 Finland 2007 France 2007 Italy 2006 Germany 1998 Denmark 2001 Denmark 2001 Germany 2002 Norway 2001 Hungary 2002 Czech R 2006 Norway 2005 Czech R 2002 Germany 2005 Sweden 2002 Sweden 1998 Netherlands 2002 Sweden 2006 Netherlands 1998 0 5 10 15 20 25 Individuals self located on the centre (%) (b) Like with the other ideological positions, the starting point in order to understand what the centre stands for is to apply the debate about which component prevails when the individual locates himself/herself on the scale. That is, following the logic explained in the previous subsection, one can locate herself on the centre because, 7
first, it can have a party of reference close to her position (a centrist party) or, second, because this position is the expression of different centrist values or attitudes. The reasons why individuals locate themselves on the centre are not, however, the same than the indicated above. Despite the lack of attention received by scholars, the literature has not left the issue into a blind spot and some hypotheses have been put forward that are different from the component-debate detailed above. In particular, Knutsen (1998b) suggested three hypotheses that can explain the individual centrist self-location: A genuine position: For Knutsen (1998b, 303) this is the most obvious way to interpret the centre position. The centre would be a genuine position as long as the individuals interpret this position as the mid-point between leftist and rightist positions. The left-right axis has the attribute of being a continuum, which allows for extremist positions (leftist or rightist) and some other positions in-between. Therefore, individuals that are not extremist will locate themselves on the centre because this position is between what left and right stand for. In other words, when individuals hold a moderate position between leftist and rightist values, they will locate themselves on the centre because this is precisely the space between the left and right position. This is why the notion of the centre has sometimes been associated with pragmatism or a simply equidistant position between the two extremes. A non-response position: In 1983 Lambert (1983, 143) pointed out that the central assumption [of the left-right scale] is that respondents sometimes use the midpoint of the scale when they are unsure of a scale s meaning or if they lack information about a party. A similar idea was hold by Laponce (1970) 5. Later, Converse and Pierce (1986, 128) further argued that the mid-point is an obvious selection for a person who is neutral, uncommitted, and even thoroughly indifferent to or ignorant about this generic axis of dispute. These ideas crystallised in Knutsen s hypotheses, who suggested that the centrist location could be a concealed form of nonresponse. According to this idea, people with low political sophistication/information will be more likely to place themselves on the centre position. The idea is that the centrist position may be an alternative to answering don t know or no answer. The individual, instead of avoiding a dk/na option, would choose a neutral alternative, 5 This idea was sometimes used to defend that, precisely because of the strong centrist tendency, left and right were meaningless tools for political science (Ogmundson, 1979; Ronald D. Lambert and Kay, 1986) 8
i.e., the centre position 6. An irrelevant category: Finally, Knutsen (1998b) considered that the centrist self-placement may be also an expression of the irrelevance of left and right among the population. This hypothesis departs from the idea that politics is multi-dimensional and that, beyond the commonly used left-right axis, individuals take into account different dimensions when analysing politics which are not aligned with the common left-right axis (Stokes, 1963; Castles and Mair, 1984; Middendorp, 1992; Geoffrey Evans and Lalljee, 1996; Warwick, 2002; Grendstad, 2003; Benoit and Laver, 2004; Hix, Noury and Roland, 2006; Lachat, 2009). In other words, they consider that the left-right language is outmoded and about to be overcome by a new political language (Kitschelt and Hellemans, 1990b, 214). Precisely because individuals give priority to other dimensions or because they think left and right are concepts linked to old politics, when facing the left-right self-location question, they may not have a clear idea on which position they should be located (de Benoist, 1995; Jahn, 2011). Again, following the non-response position hypothesis, individuals, rather than refusing to place themselves on the scale, will locate themselves on the centre, as they may consider this position as a sort of a neutral uncomitted point. As already highlighted by Knutsen (1998b, 304), this hypothesis differs from the non-response hypothesis in a fundamental aspect: whereas for the non-response hypothesis the centre self-location is a product of the individual s lack of political information or sophistication, the irrelevant category hypothesis stresses that political information is a necessary condition to judge whether the left-right axis is considered irrelevant or valid. Knutsen also labels the irrelevant hypothesis as evaluative don t knowers : people who are well informed about politics, and who know fairly well the conventional meaning of left and right, but consider these terms irrelevant for their own political thinking. In order to consider left and right as irrelevant one must have a pre-existing idea of what these terms stand for. Therefore, although both hypotheses lead the individuals to locate themselves on the same ideological positions, the point of departure is significantly different. In conclusion, as developed here, centrist self-placement can be the product of different factors: 1) A centrist genuine position, which is taken for individuals that have moderate values or attitudes, in-between left and right. 2) A product of the individual s preferred party position. 3) A position resulting from the lack of political 6 In a similar vein, see also Luskin (1987); Palfrey and Poole (1987) 9
information. 4) Finally, a by-product of the importance given to new dimensions and the consideration that the left-right schema is of little use to understand today s politics. 3 Research Design In order to test my hypotheses I employ the three modules of the Comparative Study of Electoral Systems 7. The three modules of the survey have been merged and, after excluding non-european and non-democratic countries, 52 elections in 21 different countries remain. The reason to stick to European countries is two-fold: First, previous research has shown that the left-right dimension has been proven to be comparable across European countries (Huber, 1989; Knutsen, 1997, 1998a; Linzer, 2011). Second, most of the research undertaken in this field has focused on European countries, which allows me to compare the results obtained here with previous analysis. In the final dataset different elections of the same country are included. However, all of them are treated separately and are not pooled into a single cluster. The reason is that each election is different and therefore may have a different impact on the individuals left-right self-placement. It is also a conventional wisdom that left-right mean different things across countries and over time (Fuchs and Klingemann, 1989; Linzer, 2011; Jahn, 2011). This implies that people from the same country can locate themselves on the centre for different reasons than citizens of other countries. Furthermore, individuals from a given country con locate themselves on the centre at time t and do it again at time t+1 but for different reasons. Therefore, by dealing with the elections separately, it is possible to explore the different hypotheses under different contextual scenarios. The dependent variable to be used here stems from the common left-right self-location question. The wording of the question in the CSES is the following: In politics people sometimes talk of left and right. Where would you place yourself on a scale from 0 to 10 where 0 means the left and 10 means the right? This question has been recoded into three categories in order to identify whether the respondent locates herself on the left (0-4), on the centre (5) or on the right 7 All the modules employed are freely available at http://www.cses.org. 10
(6-10) 8. It is important to point out two I did not group the left and the right together because, if it was the case, the leftist and rightist effects can be balanced out. At the same time, and unlike some other ideological scales commonly employed (Inglehart and Klingemann, 1976, 247), the CSES scale has a real mid-point. The ten-point self-placement scale, which is also commonly employed in other surveys, has no mid-point and therefore the centre is not possible to identify in a direct way. Because of the operationalisation of the CSES ideological scale, this problem is avoided in the analysis performed here. The dependent variable has three categories. Therefore the model employed in the analysis is a multinomial logit. The party-component hypothesis suggests that the centre category is a mirror of the preferred party placement. The most appropriate way to implement this idea is employing the quadratic proximity utility equation, which has been the cornerstone of most empirical analysis of voting behaviour. More concretely, and considering the left-right dimension, voter s utility is specified as the negative of the squared distance between the voter i and the party l (Merrill III and Grofman, 1999). U i = (V i P l ) 2, (1) where U i is voter s i s utility; V i the voter i s position on the left-right scale; P l the position of the closest preferred party as perceived by voter i; and U il is the voter i s utility for party l. In order to identify each individual s most preferred party I used the like-dislike thermometer, which is included in the three modules of the CSES. Using this variable I proceeded in three steps: First, I identified the most preferred party for each individual included in the survey (unless all the parties receive a zero on the scale) 9. Second, parties position in the ideological scale were calculated according to the subjective left-right positioning, that is, respondents ideological positioning of each party 10. Finally, the distance between individuals self-placement and the ideological position of the most preferred party was calculated following the utility equation. The logic behind utility equation is that the closer the party to an individual, the 8 I excluded those that did not answer the question. 9 Generally speaking, all the parties receive rather low values on the scale. If I restrict the position beyond which a party is considered as preferred (for instance, from 4 or 5 onwards) results still hold, although with an important loss of accuracy due to the high number of missing values. 10 More than 50% of the sample located one of the parties in the same ideological position as the respondent s, most probably due to the so-called assimilation effect. 11
more likely is this individual to vote for this party (Downs, 1957; Adams, Merrill and Grofman, 2005). Applying this notion to the party component hypothesis the logic is similar: the closer the party to the centre position, the more likely is the individual to locate herself on the centre 11. In other words, people located themselves on the centre because his/her preferred party is closer to the centrist position. The non-response hypothesis pointed out that centrist self-placement is a consequence of the lack of political information. The CSES modules provide three political information questions tapping individuals knowledge about different political issues. These questions were administered differently in each country, as each country sample referred to a variety of national political issues. Using this question, an index has been built, which ranges from 0 (all the answers are correct) to 3 (all the answers are incorrect). The irrelevant hypothesis defended that the centrist self-placement is a consequence of the emergence of new dimensions that are not correlated with the left-right axis. In this article the so-called new politics cleavage will play such role, particularly concerns about the environment, global peace and democracy. In order to capture these concepts, I used the following question included in the second and the third module of the CSES: What do you think has been the most important issue facing [country] over the last [number of years that the last government was in office] years?. For each country-election included in the dataset, I created a dummy variable identifying whether the respondent chose an issue related to the environment, peace or general concerns about the quality of democracy. For instance, in the 2001 Danish elections those respondents that answer that the most important problem faced by the country was Democracy/Human rights, Ecology, Environment and Development Aid were classified as having postmaterialist values. In the 2003 Finish answers such as Peace, Democracy and Nature conservation were also coded as postmaterialist. These values are generally considered a central part of the postmaterialist thinking and, although some studies have shown that in some countries individual s position on postmaterialist issues correlates with the left-right position (Kitschelt and Hellemans, 1990a; Neumayer, 2004), recent research provides strong evidences 11 I am aware that this idea breaks one of the Downsian assumptions (Stokes, 1963), that is, that individual s position on the scale is exogenous to parties position. Therefore, if the most preferred party moves away from the centre, it is assumed that individuals will follow it if the party component is crucial. It is not the purpose of this article to make any causal claim between individuals and parties positions, but to test the relationship between both for centrist individuals. 12
that these issues emerge as a separate dimension (Bennulf and Holmberg, 1990; Blake, Guppy and Urmetzer, 1997; Seippel, 1999; Meguid, 2005; Carter, 2006; Green-Pedersen, 2007). A simple correlation proves that there is no relationship between leftright positioning and the postmaterialist variable (-0.0092 and not statistically significant). Moreover, there is a limitation with the data, as there is no other dimension that would have allowed to properly check its correlation with the left-right axis. Finally, precisely because of the lack of different issues, the genuine hypothesis cannot be tested directly. Recall that this hypothesis considers the centre as a moderate position whereby individuals that see themselves in a position between left and right end up self-placing themselves. Even though this hypothesis cannot be tested in a direct empirical way, it is possible to test it indirectly. That is, if the rest of the hypotheses do not hold, it will imply that the genuine hypothesis is confirmed. If the analysis reveals that the centrist self-location is neither a product of the lack of political sophistication nor of the party component or postmaterialist values, it means that it is considered for individuals as a genuine position. The empirical analysis is divided into two parts. In the first part I perform an aggregate analysis in order to explore whether there is any aggregate relationship between the variables of interest. In the second part I perform a multinomial logit model (being in the right is the reference category). Following King, Tomz and Wittenberg (2000), in order to demonstrate the size of the effects obtained in the analysis, I calculated first differences, which illustrate how the probability to be on the centre changes when the three variables of interest are altered while holding other variables at their means. 4 Data Analysis In order to explore the relationship between centrist self-placemenet and the hypotheses presented above, I first perform different aggregate bivariate relationship. The following scatterplots show the relationship between the percentage of individuals self-located on the centre and the level of polarisation 12, the percentage 12 The index of polarisation is taken from Hazan (1998). It is expressed as follows: N i=1 p i(x i x) 2, where N is the number of parties in the system, p i is the proportion of votes won by each party, x i is the respective ideological position of each party and x is the weighted system mean, that is, the sum of the product of each party s proportion of the vote and its position on the left-right scale. 13
of people that did not know or did not answer the self-placement question and the percentage of individuals that mentioned as salient postmaterialist issues. In a polarised scenario, political alternatives are far away from the centre, whereas in a convergent scenario, parties are clustered around the centre 13. If centrist self-placement depends upon a party of reference, it is reasonable to think that centrists will have more closer options in a convergent scenario than in a polarised one. Following this logic, there should be a relationship between the lack of polarisation and the number of citizens located on this ideological position. Figure 2(a) shows that there is association between both variables, but it is not very strong (bivariate correlation is -0.19 and not significant). When polarisation increases, less citizens locate themselves on the centre, although the trend line is almost flat. Figure 2(b) plots the percentage of citizens that locate themselves on the centre and the percentage that did not answer or did not know the question. Following the non-response hypothesis, citizens that do not know where to locate themselves, can end up on the centre. This graph intends to show whether there is a trade-off between both processes. It is perhaps possible that in some contexts the pressure to answer the question is higher than in other contexts (or because of the negative connotations attached to not answering the question). Conversely, when this pressure is not present, citizens will be able to freely locate to where they think. Surprisingly, the relationship is the opposite as expected. Figure shows that the percentage of centrist individuals is positively associated with the percentage of dk/na individuals. Therefore, in those countries where more people located themselves on the centre, it also tends to exist a higher percentage of centrist citizens (correlation equals 0.45 and significant at the 95% level). Finally, the relationship between centrist self-placement and postmodernist values is also very weak (correlation is -0.19 and not significant). Moreover, it works in the opposite direction: when the percentage of postmodernists goes up, the percentage of centrist individuals goes down. 13 The literature has provided a counter-argument as to what is the relationship between polarisation and left-right conceptions. Freire (2008) argues that clarity of political alternatives (polarisation) helps citizens anchoring their position on the left-right axis. It is not the goal of this part to test a causal claim or alternative hypotheses regarding polarisation and left-right self-placement. The purpose is mainly to check the existance of bivariate relationships in order to provide insights to study the individual micro-foundations of centrist self-placement. 14
Figure 2: Scatterplots between centrist self-placement, polarisation, dk/na option and postmodernism Centrist individuals (%) 15 20 25 30 35 40 45 50 Centrist individuals (%) 15 20 25 30 35 40 45 50 2 4 6 8 Polarisation index (a) Polarisation 10 20 30 40 Dk/Na individuals (%) (b) Don t Know/No answer Centrist individuals (%) 15 20 25 30 35 40 45 50 10 20 30 40 50 60 Postmodern individuals (%) (c) Postmodernism The next part investigates for each country the impact of party proximity, political sophistication and postmaterialist values. Thus, for each of the 52 country-elections, I ran a multinomial logistic model (rightist category as the reference category). Each model included different sociodemographic controls: age, gender (1, Male; 0, Female), education (from 1, no education, to 8, undergraduate completed) and income (1, loquest quintile; 5, highest quintile). In order to have a direct interpretation of the three variables of interest, I show the first differences for each variable and each country 14. Table 2 contains the first differences in the three key independent variables and the associated estimated difference in the probability to be on the centre position. The change in the probabilities is calculated when political sophistication or party proximity increase one standard deviation from the mean. In the case of postmodernist values, first 14 Results of the 52 multinomial logit models are available upon request. 15
differences are a result of individuals giving priority to these values versus those that did not. 16
Table 2: First differences Country-election Factors F.D. L.B. U.B. Country-election Factors F.D. L.B. U.B. 17 Switzerland 1999 Czech Republic 1996 Germany 1998 Denmark 2001 Spain 1996 Spain 2000 Proximity 0.005-0.001 0.01 Proximity 0.02 0.01 0.03 P.sophistication 0.05 0.02 0.08 Ireland 2002 P.sophistication -0.0002-0.03 0.02 Postmaterialism - - - Postmaterialism 0.01-0.03 0.06 Proximity -0.01-0.03-0.01 Proximity 0.02 0.003 0.04 P.sophistication 0.02-0.00 0.05 Iceland 2003 P.sophistication 0.01-0.02 0.03 Postmaterialism - - - Postmaterialism 0.01-0.04 0.08 Proximity 0.005-0.004 0.01 Proximity -0.03-0.05-0.01 P.sophistication -0.01-0.02 0.01 Italy 2006 P.sophistication 0.06 0.02 0.10 Postmaterialism - - - Postmaterialism -0.05-0.11 0.01 Proximity 0.03 0.01 0.04 Proximity -0.001-0.01 0.003 P.sophistication 0.01-0.01 0.03 Netherlands 2002 P.sophistication 0.03 0.01 0.05 Postmaterialism - - - Postmaterialism 0.01-0.03 0.05 Proximity 0.001-0.01 0.01 Proximity 0.04 0.02 0.05 P.sophistication 0.02-0.03 0.07 Norway 2001 P.sophistication 0.03 0.04 0.01 Postmaterialism - - - Postmaterialism -0.02-0.06 0.01 Proximity -0.01-0.02 0.001 Proximity -0.002-0.01 0.004 P.sophistication 0.04 0.01 0.07 Poland 2001 P.sophistication 0.05 0.01 0.08 Postmaterialism - - - Postmaterialism -0.03-0.09 0.03 Continued on next page
18 UK 1997 Hungary 1998 Iceland 1999 Netherlands 1998 Norway 1997 Poland 1997 Portugal 2002 Proximity 0.03 0.01 0.04 Proximity -0.01-0.02-0.005 P.sophistication 0.006-0.01 0.03 Portugal 2002 P.sophistication -0.01-0.03 0.03 Postmaterialism - - - Postmaterialism - - - Proximity -0.01-0.03-0.005 Proximity 0.0002-0.004 0.005 P.sophistication 0.03-0.01 0.06 Portugal 2005 P.sophistication 0.001-0.02 0.02 Postmaterialism - - - Postmaterialism 0.05 0.005 0.09 Proximity 0.04 0.02 0.05 Proximity -0.002-0.03 0.03 P.sophistication -0.02-0.04 0.01 Romania 2004 P.sophistication -0.004-0.01 0.0003 Postmaterialism - - - Postmaterialism 0.08-0.02 0.18 Proximity -0.004-0.01 0.0001 Proximity 0.01-0.004 0.03 P.sophistication -0.004-0.02 0.01 Slovenia 2004 P.sophistication 0.04-0.004 0.08 Postmaterialism - - - Postmaterialism -0.02-0.12 0.09 Proximity 0.006-0.03 0.01 Proximity -0.01-0.02 0.005 P.sophistication 0.02-0.004 0.04 Sweden 2002 P.sophistication 0.06 0.03 0.09 Postmaterialism - - - Postmaterialism -0.01-0.07 0.04 Proximity -0.01-0.02-0.004 Proximity 0.001-0.01 0.01 P.sophistication 0.03 0.01 0.06 Austria 2008 P.sophistication 0.02-0.01 0.05 Postmaterialism - - - Postmaterialism - - - Proximity -0.01-0.02-0.004 Proximity 0.02 0.01 0.02 P.sophistication 0.003-0.03 0.04 Switzerland 2007 P.sophistication 0.06 0.04 0.07 Continued on next page
19 Romania 1996 Slovenia 1996 Sweden 1998 Bulgary 2001 Switzerland 2003 Czech Republic 2002 Germany 2002 Postmaterialism - - - Postmaterialism -0.06-0.09-0.02 Proximity 0.01-0.004 0.02 Proximity -0.01-0.02-0.01 P.sophistication 0.006-0.03 0.05 Czech R 2006 P.sophistication 0.04 0.01 0.06 Postmaterialism - - - Postmaterialism -0.04-0.09-0.001 Proximity 0.01-0.005 0.02 Proximity 0.01-0.002 0.02 P.sophistication -0.01-0.04 0.02 Germany 2005 P.sophistication 0.02-0.001 0.01 Postmaterialism - - - Postmaterialism - - - Proximity -0.0003-0.005 0.005 Proximity 0.01-0.001 0.02 P.sophistication 0.03-0.001 0.06 Germany 2009 P.sophistication 0.02-0.0005 0.05 Postmaterialism - - - Postmaterialism -0.11-0.18-0.01 Proximity -0.01-0.01-0.0003 Proximity 0.01-0.001 0.02 P.sophistication 0.03-0.02 0.03 Finland 2007 P.sophistication 0.04 0.01 0.07 Postmaterialism - - - Postmaterialism -0.04-0.09 0.01 Proximity 0.004-0.01 0.01 Proximity 0.002-0.004 0.01 P.sophistication 0.02-0.001 0.05 France 2007 P.sophistication 0.001-0.02 0.02 Postmaterialism -0.02-0.10-0.06 Postmaterialism -0.01-0.05 0.03 Proximity -0.01-0.02-0.0001 Proximity 0.01-0.002 0.03 P.sophistication 0.04 0.01 0.07 Ireland 2007 P.sophistication 0.006-0.03 0.04 Postmaterialism - - - Postmaterialism - - - Proximity 0.01 0.006 0.03 Proximity 0.08 0.05 0.11 P.sophistication 0.01-0.02 0.04 Iceland 2007 P.sophistication 0.03 0.004 0.06 Continued on next page
20 Denmark 2001 Spain 2004 Finland 2003 France 2002 UK 2005 Hungary 2002 Postmaterialism 0.04-0.06 0.18 Postmaterialism -0.03-0.08 0.02 Proximity 0.02 0.003 0.03 Proximity 0.0001-0.01 0.01 P.sophistication -0.007-0.03 0.01 Iceland 2009 P.sophistication -0.01-0.04 0.03 Postmaterialism 0.06-0.01 0.16 Postmaterialism 0.02-0.03 0.08 Proximity -0.01-0.02-0.002 Proximity 0.03 0.01 0.04 P.sophistication 0.02-0.01 0.05 Norway 2005 P.sophistication 0.04 0.02 0.06 Postmaterialism - - - Postmaterialism -0.01-0.04 0.03 Proximity 0.01-0.001 0.03 Proximity -0.01-0.01 0.001 P.sophistication 0.04 0.01 0.07 Poland 2005 P.sophistication 0.05 0.03 0.08 Postmaterialism 0.02-0.05 0.09 Postmaterialism - - - Proximity 0.001-0.01 0.01 Proximity -0.004-0.01 0.005 P.sophistication 0.03-0.01 0.06 Poland 2007 P.sophistication 0.03 0.004 0.06 Postmaterialism 0.001-0.07 0.07 Postmaterialism -0.002-0.05 0.04 Proximity 0.04 0.01 0.07 Proximity -0.004-0.01 0.005 P.sophistication 0.03-0.01 0.08 Portugal 2009 P.sophistication 0.03 0.004 0.06 Postmaterialism -0.002-0.10 0.11 Postmaterialism -0.002-0.05 0.04 Proximity -0.01-0.01 0.001 Proximity 0.003-0.01 0.01 P.sophistication 0.04 0.01 0.08 Sweden 2006 P.sophistication 0.02-0.01 0.05 Postmaterialism 0.04-0.01 0.09 Postmaterialism 0.05-0.004 0.11
Just eyeballing the previous table, a clear-cut evidence arises: the effect of each component on the probability to be located on the centre varies across countries as well as over time. There is not a single component that prevails over the others in a clear way. At the same time, it emerges as a regularity that in each of the countries only one factor drives citizens to locate themselves on the centre. Only in the case of Norway 2001, Poland 1997, Switzerland 2007, Iceland 2007, and Norway 2005 two components are significant 15. Party proximity emerges as the most important factor that leads citizens to place themselves on the centre of the scale. It is significant in nineteen cases. The party component is significant and relevant when centrist citizens see that the closest preferred party approaches the centre position. That is, the closer the party to the centre, the higher the likelihood to be located on the centre because of the party s preferred position. Political sophistication is significant in sixteen cases. When political sophistication increases one standard deviation from the mean, individuals are more likely to place themselves on the centre relative to the right. As for postmodernist values, it is only significant in three cases and in two it goes in the opposite direction. Despite these findings, several important nuisances should be highglighted: First, it is important to mention that in sixteen cases any of the first differences was significant, which indirectly confirm the genuine hypothesis. Second, even though proximity and political sophistication are components that, together, explain centrist self-placement in most of the countries, the size of the effect is generally small. In order to clearly show this pattern I plot the first differences coefficients against the percentage of centrist individuals in each country. If party proximity and political sophistication matter, differences across countries in centrist self-placement will be explained by a larger or smaller effect of these factors. 15 In line with previous findings, the analysis does not show either a different pattern for Post-Communist countries (McAllister and White, 2007). 21
Figure 3: Scatterplots between centrist self-placement and the effect of party proximity and political sophistication Centrist individuals (%) 15 20 25 30 35 40 45 50 Centrist individuals (%) 15 20 25 30 35 40 45 50 0.2 0.1 0.0 0.1 Political sophistication (a) Political sophistication 0.15 0.10 0.05 0.00 0.05 0.10 0.15 Party proximity (b) Party proximity Figure 3(a) and 3(b) confirm the previous intuition. When the effect of political sophistication increases, the percentage of centrist individuals tends to go down, although the line trend is almost flat. As for party proximity, results go in the opposite direction: when the effect of party proximity increases, the percentage of centrist individuals goes down. Although this seems counterintuitive, it is reasonable if we take into account that this analysis is static rather than dynamic. As mentioned, the endogenous nature between party position and individuals position (Adams, Merrill and Grofman, 2005) makes the previous analysis very noisy. Most probably, people adapt their position when parties move along the axis, something difficult to grasp with the data used here. However, even if the later is true, it might be the case that individuals post-rationalise party movements. That is, if the centrist individual has the closest preferred party far away from its position, she will not use a partisan cue to justify her position and will present herself as a simple moderate. However, if this party moves closer to the centre, now this same individual can justify her position on the party s new positioning. 5 Conclusion and implications In most of the European countries the existence of the centre is not only a mathematical term, but also a political one. Although this afirmation seems obvious, the centre has received much less atention thah its ideological counterparts. Despite its quantitative and qualitative importance, the question of why citizens place 22
themselves on the centre still remains unresolved. Some studies have raised some hypotheses and highlighted the necessity to focus on this position because individuals can locate on it for different reasons than on other ideological positions. This centrist excepctionalism has been concreted in four hypotheses: the party-component hypothesis, the non-respondent hypothesis, the irrelevant hypothesis and the genuine hypothesis. Findings in this article suggests that the centre position is more of a genuine position than what some previous scholars had considered. It is true that the effect of each of the components varies across countries and that in some cases party-proximity and the lack of political sophistication play a more important role. However, the effect of both variables is rather small and sometimes contradictory. This article therefore achieved very similar results than the ones obtained by Knutsen (1998b) or Torcal (2011) for the Spanish case. They both concluded that the centre is, above all, a genuine position. This finding is also consistent with previous research about voting behaviour that shows that centrists individuals make more sophisticated judgements when deciding his vote preference (Tomz, 2008). At the same time, this article does not find any different systematic pattern for the Post-Communist countries, confirming previous research on this field (Bohrer and Radcliff, 2000; Jou, 2011). Even though findings on this article represent a step forward towards our understanding of the centre position, there are still some issues that future studies will have to deal with. The relationship between the party component and the centrist position is especially important. The endogeneous nature of such a relationship should be tackled with experimental designs or panel data, with which the causal mechanism can be really isolated. At the same time, the different components included in the analysis might be multicolineal or endogenous. In such a situation, structural equation modelling will be more appropiated, although another important difficulty will emerge: the model will not probably be identified because it will be theoretically impossible to derive a unique estimate of each parameter. In any case, the centre is a crucial position to understand how parties behave (Enelow and Hinich, 1984; Adams, Merrill and Grofman, 2005). This research sheds some light on this obscure position by testing different hypotheses that may have led to talk about the centrist excpetionalism. However, findings indicate that the centre is what the term indicates: a space between left and right, where citizens with values or opinions in-between end up locating themselves. 23