What is the relation between System Justification and voting behaviour? : an experimental survey in The Netherlands

(1)

What is the relation between System Justification and voting behaviour?

An experimental survey in The Netherlands

Masters thesis political theory and political behaviour Name: Koenraad Wiering (10674640)

Supervisor: Gijs Schumacher Second reader: Marc van de Wardt University: University of Amsterdam

(2)

1

Abstract

This study presents the significant correlation between the level of system justification and conservative voting behaviour. According to the System Justification Theory, people have a general ideological motive to justify the existing structures in a social system. By means of this gives System Justification Theory an explanation for the conservation of the inequalities in status quo and the absent of change to a more equal society. The data of the experimental survey shows a significant positive correlation between the level of system justification and ideological conservative voting (H1) (r=0.30) and voting for current government parties (H2) (r=0.46). No support is found for H3, for which a positive relation between the level of system justification and the chance to choose the same as last elections was expected. These results might imply that SJ, by means of the positive relation with conservative voting behaviour, is partly responsible for absents of change in the current inequalities in The Netherlands.

(3)

2

Index

Introduction

In the latest report of the Netherlands Institute for Social Research (SCP) on the topic of ‘how the Dutch society is doing’, they claim: “The citizens in our country still have to deal with

many differences and inequalities in their opportunities to participate in this prosperity and to realize their individual aspirations and interests. Sometimes these differences are desirable or inevitable, but sometimes they result in unequal opportunities and outcomes that are considered unnecessary and unjust” (Sociaal en Cultureel Planbureau, 2015. p.7). These

unnecessary and unjust inequalities are nowadays still a matter of fact in the Netherlands. The same report states that over the last decade the income inequalities in the Netherlands are not decreasing and act stable (Sociaal en Cultureel Planbureau, 2015). However, when you look at the income inequalities from the beginning of the 90’s you see that the inequalities even slightly increased over the years (Vrooman & Wildeboer Schut, 2015).

An underlying question where many social scientists have worked on is: why do

disadvantaged people tolerate the unjust system instead of doing everything in their power to change the system (Jost & Thompson, 2000; Frank, 2004; Eyerman & Jamison, 1991;

Neckerman, 2004)? The ‘System Justification Theory’ (SJT), first published by Jost & Banaji (1994), gives a possible explanation for the conservation of the status quo and the absence of change in the current inequalities. According to the SJT, there is a general ideological motive to justify the existing structures in a social system. SJT proposes that people actively defend and bolster existing social arrangements, often by denying or rationalizing injustice and other problems, even when doing so comes at the expense of their personal and group interest (Jost, Nanaji, and Nosek, 2004; Jost, Burgess, & Mosso, 2001; Jost & Hunyady, 2005; Kay et al., 2007; Van der Toorn & Jost, 2014). High levels of System Justification (SJ) may explain the absent active progressive behaviour of disadvantaged people.

Following Jost, Glaser, Kruglanski & Sulloway (2003), it is expected that in a democratic system the direct influence of the SJT can be seen in voting behaviour at elections. At the national election in The Netherlands, the electorate has the opportunity to choose which party will represent their vote in the parliament for the next four years. Usually there is a choice of several parties, each having their own position and opinion towards how society is

functioning including subjects as inequality. With a vote people express their preference for the party they want to be represented by in the parliament. In this study the influence of the SJ motive on voting behaviour is researched in order to get a better understanding of the absence of change in the current inequalities.

(6)

5

Central research question

The direct impact of SJ on voting behaviour has never been researched before and is the central subject in this study. In earlier SJ literature, a lot of research is done to show what factors cause SJ (e.g. Jost & Banaji, 1994; Kay & Friesen, 2011; Van der Toorn et al, 2015). This literature is included in the theoretical framework for this study, which focuses on the effect of SJ, and more precisely the impact of SJ on voting behaviour in The Netherlands. The central question in this research is:

“What is the influence of System Justification on voting behaviour in The Netherlands?”

Answering this question will provide a better understanding of causes for inequalities in societies. The SJ motive might influence voting behaviour of citizens and ensures

consolidation of the status quo and absence of change in inequalities. The application of SJT has been studied mostly in the U.S., so far, and has been studied less often in Europe (e.g. Van der Toorn, Berkics & Jost, 2010; Jost et al, 2012). With the application of the SJT in a study in The Netherlands is sought to broaden and deepen the social psychological

understanding of cognitive and motivational factors that contribute to the legitimation of the status quo. The theories on SJ and voting behaviour have not yet been combined in a study. By combining the SJT and theory on voting behaviour the following expectations occur. It is expected that SJ has a positive relation with conservative voting behaviour. Various researches (e.g. Jost & Hunyady, 2005; Jost, Glaser, Kruglanski & Sulloway, 2003) describe that political conservatism has a strong positive impact on the level of SJ. Given this relation, it is expected that [H1] the level of SJ has a positive relation with propensities to vote for

conservative parties. High levels of SJ increase the change to justify the status quo (e.g. Jost

& Banaji, 1994; Jost & Hunyady, 2005). Since the current government parties are part of the status quo it is expected that [H2] the level of SJ has a positive relation with propensities to

vote for government parties. The justification of the status quo might also result in making the

same choice as in the previous parliamentary election. Therefore it is expected that [H3] the

level of SJ has a positive relation with the chance to make the same choice as in the last parliamentary election.

These hypotheses are explained in depth at the end of the theory section that follows next. In the theory section an insight is given on the theory of SJ, conservatism, voting behaviour and specific voting behaviour in The Netherlands. In the ensuing method section, it is explained how the hypotheses are researched and how the experimental survey is set up. The outcomes of the experimental survey are presented in the following results section. The consequences and implications of the results are given in the conclusion section. The discussion in the

(7)

6 conclusion section puts the findings from this study in context of the already existing

(8)

7

Theory

System Justification Theory (SJT)

A potential explanation for the conservation of the current inequalities in society is found in SJT. According to SJT, people adopt ideologies and belief systems that serve as excuses and justifications for existing social, economic, and political arrangements to make them feel better about the status quo (Jost, Banaji, & Nosek, 2004). System justifying ideologies include belief systems that rationalize inequality and emphasize on the use of stereotypes that fulfil a remedial function (Kay & Jost, 2003). SJT suggest that people are motivated to defend, bolster and rationalize the social system that affect them because doing so serves basic epistemic, existential and relational needs (Jost, Burgess & Mosso, 2001). This is reflected in believing that the current social system is the way a social system should be. The extent to which people defend and bolster the existing social arrangement vary between individuals, groups and systems. The SJT doesn’t imply that people always experience the status quo as fair and just (Feygina, Jost & Goldsmith, 2010). However, several studies show evidence that giving people the opportunity to justify the system, leads them to feel better and more satisfied, feeling more positive emotions and fewer negative emotions (e.g., Jost, Wakslak & Tyler, 2008).

The term “system” is referred to as any social, economic, or political arrangement in which multiple individuals or groups are embedded (Wakslak, Jost & Bauer, 2011). This can range form small-scale system such as family up to large systems as nation states, which is the case in this study.

To understand how SJ works, it is useful to question what it is that motivates other people to advocate the unequal redistribution of valued resources? Research by Montada, Schmitt & Dalbert (1986) and Montada & Schneider (1989) suggest that the emotional reactions of individuals when they are confronted with inequality are strong predictors of whether or not they will commit to helping the disadvantaged. Certain forms of distress, especially moral outrage, are important motivators for these actions to help the underprivileged. System-justifying ideologies reduce those emotional distresses. People that either actively endorse or are passively exposed to ideologies like SJ, should be less likely to take steps to eliminate the source of that distress and directly redress the inequality (e.g. Lerner, 1980; Walster et al., 1973). As a result, the effects of system justification may well lead to acceptance and a withdrawal of support for social change and the redistribution of resources (Jost & Hunyady, 2005). The SJ motive helps to reduce uncertainty, manage threats and may facilitate the sense that values and attitudes are socially shared. These characteristics of the SJT are closely related to cognitive dissonance (Festinger, 1957), which is explained in the next section.

(9)

8

SJT compared with cognitive dissonance theory

More characteristics of SJT will be made clear by means of making the distinction with the cognitive dissonance theory. While the justifying and rationalising mechanisms of SJT have similarities with cognitive dissonance theory and cause a strong relation between both (Jost, Pelham, Sheldon, & Sullivan, 2002), they do differ in three fundamental ways (Jost & Hunyady, 2002).

Cognitive dissonance theory (e.g. Festinger, 1957) is one of the most prominent social psychological theories of justification and rationalization processes. Festinger’s cognitive dissonance theory suggests that we have an inner drive to hold all our attitudes and beliefs in harmony and avoid disharmony. Cognitive dissonance refers to a situation involving

conflicting attitudes, beliefs or behaviour. These dissonances produce a feeling of discomfort leading to an altering in attitudes, beliefs or behaviour to reduce the discomfort and restore balance.

The first distinction between both theories is that the cognitive dissonance theory focuses on the justification of the ego, where the central aim in SJT is the justification of the system. The reduction of dissonance in cognitive dissonance theory is driven by the desire to preserve a positive image of the self, following acts of hypocrisy (Anonson, 1992; Greenwald & Ronis, 1978). In SJT people reduce ideological dissonance and defend the legitimacy of the system in order to maintain a positive image of the system. The legitimacy of the system may even come at the expense of a positive self-image or a positive group image. Second difference between both is the distinction between personal responsibilities. Most cognitive dissonance theories assume that people must feel personally responsible for the aversive consequences of an action in order to justify it (e.g. Cooper & Fazio, 1984). Instead, SJT suggests that people justify the status quo, even when they themselves have no direct responsibility for it (Kay et al., 2002). The last difference is that the cognitive dissonance theory aims at cognitive consistency (Abelson, Aronson, McGruire, Newcomb, Rosenberg & Tannenbaum, 1968), while the SJT aims to see the system as fair and just, even if such beliefs actually create dissonance, ambivalence or conflict (Jost & Burgess, 2000).

(10)

9

Conservatism

The ideology of conservatism has a long history, in which the meaning and definition changed over time. Historical and cultural variation in political systems affects the meaning of conservatism. In this study we follow the conceptual definition used by Jost, Glaser, Kruglanski and Sulloway in their article ‘Political conservatism as motivated social cognition’ (2003). This conceptual definition for the concept conservatism is a commonly used approach in the field of SJT.

Jost, Glaser, Kruglanski and Sulloway build on the approach of Abric (2001) in which political conservatism has a stable definitional core and a set of more malleable, historically changing associations, also known as ‘secondary issues’ (Huntington, 1957). The two core aspects of conservative ideology are: resistance to change and acceptance of inequality. The first core aspect of conservatism comes forward in different definitions of conservatism, such as: “the disposition and tendency to preserve what is established; opposition to change” (Neilson, 1958, p.568) or “the disposition in politics to maintain the existing order” (Morris, 1976, P. 312). The second core issue of the definition of conservatism concerns the

acceptance of inequality. Giddens (1998), following Bobbio (1996), wrote, “One major criterion continually reappears in distinguishing left from right: attitudes toward equality. The left favours greater equality, while the right sees society as inevitably hierarchical” (p.40). This characterization is consistent with many historical and political definitions of

conservative and right wing ideology (Muller, 2001).

Even though the two core issues of conservatism are strongly related to each other, they don’t necessarily always go hand in hand. There are historical examples in which the two are distinguishable and not related to each other. For instance in the case of right-wing

revolutionaries that aren’t resisting inequalities but also enhance change. Despite exceptions like the one mentioned, the two core aspects of conservatism are generally psychological related to one another, for most of the people, most of the time (Muller, 2001). However, results of Jost et al (2007) suggest that the two “core” aspects of conservatism may be triggered by distinctive, sometimes related cognitive-motivational processes. According to Federico, Ergun and Hunt (2014), the need of cognitive closure is more strongly linked to political conservatism through resistance to change than opposition to equality.

In this study the definition of conservatism that is most appropriate in the field of SJT will be used. This way, it is ensured that the junction to the existing literature on SJT and the specific focus of this study are both maintained. This study, especially for hypotheses 2 and 3, focuses more on the core aspect of conservatism ‘resistance to change’.

(11)

10

Voting behaviour

At the parliamentary election in The Netherlands, the electorate has the opportunity to choose which party will represent their vote in parliament for the next four years. With a vote people express their preference by which party they want to be represented in parliament. Different theories presume to explain the vote choice and what causes voting motives. The spatial and valence theories are often used explanations for electoral choice (Sanders, Clarke, Stewart & Whitely, 2011).

The spatial theory is primarily inspired by the work of Downs (1957), which asserts that voters assess their own and competing parties positions on one or more ideological or policy dimensions. By selecting the party that is closest to them in the dimensions space, voters maximize their expected utilities. Spatial analysts may differ in the precise way they measure and define the position and proximities between individuals and parties. The main principle of the spatial theory remains the utilisation of the relative position of voters and parties on the relevant dimensions.

The valence theory is raised after critique by Stroke (1963) on the spatial models. The valence theory does not deny that individuals and parties can be located at different positions on various policy and ideological dimensions. For the valence theory these differences are much less important than party’s rival attractions in terms of their perceived ability to handle the most serious problems the country is facing. Voters maximize their utilities by choosing the party they think is most capable of delivering policy success in the areas that concern them the most. So what matters, are the comparative capabilities and potential performances of parties when they attempt to deal with the important problems.

Both the spatial and the valence theory do have one important limitation: both theories are assuming that voters make rational decisions. The assumption that voting choice is a rational decision is questionable. For a long time, research on decision making has underestimated the role of affect and unconscious evaluations. The aim was primarily to develop normative models in which decision making was considered as a very rational process for maximizing expected utilities (Edwards, 1961). The spatial and valence theory are examples of this. Recently greater attention has been paid to both affective responses and unconscious processes. Affective responses have been incorporated into models of political decision making (Abelson, Kinder, Peters & Fiske, 1982; Marcus & Mackuen, 2001; Glaser & Salovey, 1998; Arcuri et al, 2008; Friese et al, 2007).

The SJ motive, which works partly unconscious, is the independent variable in a part of the analyses in this study. This study focuses mainly on unconscious motives instead of the

(12)

11 conformal, rational theories. The aim of this study is partly to expand the described non-rational and unconscious theories.

Voting behaviour in The Netherlands

For the last 20 years, the elections in The Netherlands have been the most volatile and capricious of West Europe (Mair, 2008). At the national elections of 2002, up to 31% of the seats in the parliament changed from owning party (Van der Meer et al, 2012). At the elections of 2010 the percentage of changing seats was 24. Similar changes and instabilities have been very rare in Europe since the Second World War, and usually only occur in case of constitutional changes or collapse of a party system.

For Dutch citizens, loyalty and support for political parties during the latest 20 years are something that is not naturally and unalterably, which makes the Dutch voters hard to predict these days (Van der Meer et al, 2012). This is remarkable when you look at the

predictableness of the Dutch voters until the 80s. Until then, the Dutch voters were one of the most predictable of Europe. Before the 60s the results of the last elections were mostly a good predictor for the next and didn’t change much. Until the 80s, the distributions in the society of religion and employment status determent for a big part the distribution of seats at the

elections (Anderweg & Irwin, 2005; Aarts & Thomassen, 2008). However, between 2006 and 2010 more than half (55%) of the Dutch citizens changed at least one time in vote choice (Van der Meer et al, 2012). Worth mentioning here is that while the citizens do change between parties, they stay loyal to a group of parties that are close to each other in terms of ideology and preferences.

At the last parliamentary elections in 2012, the liberal conservative (Krouwel, 2012) ‘VVD’ won the most seats with 26,58% of the votes (Kiesraad, 2015). Together with the social democrats of the ‘PvdA’ who got 24,84% of the votes, they had more then half seats of the parliament, enough to form a government coalition. The third biggest party at the 2012 elections was the ‘PVV’ (10,08%), which is characterised by their conservative, populist and anti-Islam statements. The socialists party ‘SP’ got 9,65% of the votes, the christen democrats party ‘CDA’ got 8,51% and the social liberals party ‘D66’ 8,03%. The orthodox protestant party ‘CU’ enhanced 3.24% of the votes and the progressive left-wing greens of ‘GroenLinks’ 2,33%. These eight parties enhanced the most votes at the last elections and cover up ± 90% of the seats in the parliament. At the time when this study was conducted, the voting polls showed that these eight parties remain the largest and still cover up to 90% of the seats but in a different distribution. On the 9th_{of May, the poll of ‘Peilingwijzer’ (2016), which combines}

the scores of 5 pollsthe following percentages: VVD: 17,2, PvdA: 7,4, PVV: 20,5, SP: 10,3, CDA: 11.0, D66: 10,1, CU: 4,7, GroenLinks: 8,4.

(13)

12

Coherence of SJT, conservatism and voting behaviour

One of the reasons the SJT was developed is to investigate how and why people accept and maintain the social system that affects them (Jost & Hunyady, 2005). The failure of self-interest models to explain ideology and voting behaviour has led political researchers to search for better explanations. In the current study it is assumed that the conservative ideology, like all other belief systems, is adopted partly because it satisfies some

psychological needs. This does not mean that the conservative believes are pathological or necessary irrational, false or unprincipled (Jost, Glaser, Kruglanski & Sulloway, 2003). Following Kruglanski (1999), most human beliefs are subjectively rational in the sense of being deduced from a set of premises to which believers agree. In this sense, any given person’s conservative voting behaviour may well be principled in that it is related logically or psychologically to their observations, values and beliefs. The attachment of the principles and reasoning is an on-going process that is influenced by the environment (Hastorf & Cantril, 1954; Kunda, 1990). The possibility that SJ influences voting behaviour is considered as a matching process that takes place, whereby people adopt a ideological belief system (in this case conservatism) that is most likely to satisfy their psychological needs and motives. In a review article of Jost, Glaser, Kruglanski & Sulloway (2003) conformation is shown that the holding of politically conservative attitudes can be predicted by a set of interrelated epistemic, existential, and ideological motives, which include the system justifying motive. In their article they constructed an integrative model for political conservatism as motived social cognition, which can be found in appendix 1. The model exists of three following parts. First, environmental stimuli such as fear, threat and uncertainty influence the social cognitive stimuli. Ideological motives are one of the three social-cognitive motives that form the second part, including SJ. SJ as part of the three social-cognitive motives influences the last part, political conservatism that emerges in resistance to change and endorsement of inequality. This relation between SJ and political conservatism is fundamental for the formulated hypothesis in this study.

However, the evidence for this relation presented in the review article is only based on substantial archival historical research. These researches suggest that people are more likely to turn to authoritarian leaders and institutions of security, stability, and structure during times of societal crisis (e.g. Doty, Peterson, & Winter, 1991; McCann, 1997; Peterson et al., 1993; Rickert, 1998; Sales, 1972, 1973). The current research aims to broaden and deepen this relation by focusing on the social psychological understanding of cognitive and motivational factors that contribute to the legitimation of the status quo through voting behaviour.

(14)

13 Kay, Jimenez & Jost (2002) have emphasized the role of justification and the post hoc

rationalization of someone’s own voting preferences or behaviour. According to them, SJT implies that people rationalize both their own attitudinal/behavioural commitments, and anticipate to outcomes for which they are not responsible. A consequence of SJ, is the increased perceived legitimacy and stability of the authorities and institutions (Jost & Hunyady, 2005). This causes trust in, and approval of the government.

(15)

14

Hypotheses

The literature above showed the strong relation between the SJ and conservatism. Following this literature and these theories, hypotheses are formulated in which the level of SJ has a positive relation with tree different interpretations of conservative voting behaviour.

H1: The level of SJ has a positive relation with propensities to vote for

conservative parties

The presented relation between SJ and political conservatism causes the hypothesis that the level of system justification has a positive relation with ideological conservative voting behaviour. It is expected that higher levels of SJ increase the propensity to vote for a political party with a conservative ideology.

H2: The level of SJ has a positive relation with propensities to vote for

government parties

As part of the authority in the role of executive, the current government parties are

responsible for the governance of the country. Caused by the fact that government parties are part of the authorities, it is expected that high levels of SJ increase the propensity to vote for government parties. For this hypothesis, the ideology of the government parties is not determining to consider the voting behaviour conservative. The function as part of the authorities is the determining factor to consider the voting behaviour conservative. This expectation is in line with research on “outcome biases”, according to which people attribute favourable characteristics to winning candidates and unfavourable characteristics to losing candidates once the outcome is known (Allison, Mackie & Messick, 1996). It also suits with the political phenomenon “incumbent advantage”. This means in general that, because people prefer to vote for a winner, incumbents have structural advantages over challengers during elections (Powell & Whitten, 1993).

H3: The level of SJ has a positive relation with the chance to make the same

choice as during last elections

The resistance to change as part of the conservative mind-set may also result in the same voting choice as the last time someone chose. By making the same choice as during the last election, which may be voting for the same party or withhold from voting again, the choice does not change. This resistance to change is seen as an expression of conservative voting behaviour. Also for this hypothesis, the ideology of the party is not determining, while the chance to make the same choice as during the last elections is.

(16)

15

Method

This research focuses on the relation between SJ and conservative voting behaviour. The above-introduced three hypotheses are investigated in one experimental study including a manipulation part and a questionnaire on political preference. In this study the respondents are randomly assigned to one of the five groups: one control group with no manipulation and two manipulation methods with each two groups in which the level of SJ is attempted to be manipulated. Afterwards, voting propensities, past voting behaviour, the level of SJ and descriptive characteristics are measured.

More then two hundred respondents participated in this experiment via the online survey program ‘Qualtrics’. The size for this experiment is comparable with similar experimental studies that often have the same extent. The respondents derived from the social circle of the author of this study, and were approached online via social media or direct message. Next to the online way, other respondents have been approached directly to complete the

questionnaire on a tablet in the canteen of the university. The choice for these approaches was made due to time- and practical restrictions. The respondents participated voluntarily and did not get rewarded.

The system justification motivation was attempted to be manipulated directly by one of the two manipulation methods. The manipulation aimed to show a causal relation by finding differences between high and low manipulation in voting behaviour as reaction on the manipulation. In research studies on SJ, different ways are developed to manipulate the SJ motive. The two manipulation methods used in this study are the use of complementary stereotypes and system threat stimuli. For this study, the choice was made to use two different manipulation methods to enlarge the chance on finding distinct results based on the different levels of SJ.

Manipulation method: complementary stereotypes (Group A + B)

SJ motives can be manipulated by exposing participants to complementary stereotype exemplars, adapting methods by Kay and Jost (2003). In this method people are exposed to complementary exemplars as “poor but happy”. Kay and Jost presented that exposing people to complementary exemplars serve to increase perceptions that the status quo is fair,

legitimate, and justifiable, relative to exposing people to non-complementary exemplars. Experimental research by Eagly and Steffen (1984) and Hoffman and Hurst (1990) suggested that people use stereotypes as a way of rationalizing the unequal distributions of social roles. SJT builds on the theme that stereotypes arise to justify social and economic differences

(17)

16 between groups. This proposes that the same group will be stereotyped differently depending on whether it is perceived to be high or low in status (Jost & Hunyady, 2002).

In this experimental method the participants were told at first that the experiment would consists of two parts and that the subject is the individual trait rating of persons and political opinion. As part of the cover story and as manipulation component, first a vignette of a person is given after which the person may be rated on trait characteristics.

The respondents were asked to read a short vignette about a character named Joost, who was described as either poor + happy, or poor + unhappy. The research of Kay and Jost (2003) showed that exposure to complementary representations as ‘poor’ but ‘happy’ stereotypes led people to score higher on general measure of SJ compared with people who were exposed to non-complementary stereotypes as ‘poor’ but ‘unhappy’. The translated wording of the vignettes was as follows:

Joost is from a large Dutch city. He is married and has two children, has brown hair and is 1.80m tall. Joost liked to play football as a child and still closely follows the matches of his local team. Joost enjoys almost all aspects of his life [is not

particularly happy with most aspects of his life], but [and] because of his low salary he has trouble getting the bills paid and keeping food on the table. In June, Joost will be turning 41.

After the participants read the vignette they were asked to rate how likely or unlikely they thought it was that Joost was also: arrogant, attractive, charitable, content, corrupt, dishonest, ethical, funny, generous, greedy, competent, stupid, and unfulfilled. All ratings were made on 100 points scales ranging from 0 (very unlikely) to 100 (very likely). These questions are in addition to strengthening the cover story and several of these trait ratings allowed it to check the manipulation of perceived happiness of the person in the vignettes.

Manipulation method: system threat (Group C + D)

In another frequently used manipulation method participants read either that their country had reached a low point in terms of social, economic and political conditions (system threat), or that things are relatively fine (system affirmation) (Jost et al., 2005; Kay et al., 2005). Consistent with prior work on motivated reasoning (Dunning, Leunenberger & Sherman, 1995; Kunda, 1990), people tend to react against threatening information by supporting and bolstering their system through great endorsement of prevailing ideologies, preferences, and stereotypes of social groups (Lau, Kay & Spencer, 2008). In research of Kay, Jost and Young (2005) it is tested if the system-threat manipulation would temporarily decrease satisfaction with the status quo without threatening related constructs such as self-esteem and group

(18)

17 identification. Their results confirmed that this manipulation successfully threatened the perceived legitimacy of the system without threatening individual or collective self-esteem. In this experimental method the participants were told at first that subject of the survey is the different ways of memorising information and political opinions. First, as part of the cover story and manipulation component, the respondents read a short article from a local journalist about which they would have to answer questions later on. This short article either functions as system threat or system affirmation manipulation (Jost, Kivetz, Rubini, Guermandi & Mosso, 2005; Kay, Jost, & Young, 2005). In the system threat condition the article heightens SJ motives by challenging the social system in which participants belong. Participants were instructed to read the article carefully because they would be questioned later on in the survey.

In the high system threat condition the participants read a Dutch translation of:

These days, many people in the Netherlands feel disappointed with the country’s condition. Many citizens feel that the country has reached a low point in terms of social, economic and political factors. It seems that many countries in the world are enjoying better social, economic, and political conditions than the Netherlands. More and more Netherlanders express a willingness to emigrate to other countries.

In the low-system-threat condition the participants read a Dutch translation of:

Despite the difficulties the nation is facing, many people in the Netherlands feel safer and more secure these days compared to the past. Many citizens feel that the country is relatively stable in terms of social, economic, and political factors. It seems that compared with many countries in the world the social, economic and political conditions in the Netherlands are relatively good. Very few Netherlanders express a willingness to emigrate to other countries.

After the participants read the article they went on with four questions about the persons in the article. The participants were asked how satisfied they think the persons in the article were with the political, economical and safety situation in their country and how likely it is that those people would emigrate. All ratings were made on 100 points scales ranging from 0 (very unlikely) to 100 (very likely). The four questions were included to support the cover story about the different ways of memorising information. In this way it is sought to not make the participants unfulfilled and suspicions after they completed the survey, without

(19)

18

Control group E

In this control conditions the participants were told that subject of the survey is only political opinion. This group didn’t get any manipulation and only submitted the second part of the survey in which their political opinion and descriptive characteristics are questioned.

In the second part, the participants were asked for their propensity to vote for eight Dutch political parties: VVD, PvdA, PVV, D66, SP, CDA, GL & CU. At the time of the research, these parties had the most seats in the parliament and covered up to ±90% of all the seats. The propensities to vote-question aimed at tapping into voters (overlapping) preferences for each of the selected political parties (Van der Eijk, 2002; Van der Eijk, Van der Brug, Kroh & Franklin, 2006). The formulation of this question was based on the most recent formulation present in the European Election Study survey administered in the 28 EU countries. The translated formulation reads:

“We have a number of political parties in the Netherlands, each of which would like to get your vote. How probable is it that you will ever vote for the following parties? Please answer on a scale where ‘0’ means “not at all probable” and ‘100’ means “very probable””.

The Dutch political parties are classified as either conservative (VVD, PVV, CDA & CU) or not conservative (PvdA, D66, SP & GL) following the mapping of Kieskompas (2012). These political parties are corresponding classified in a book of one of the scientists connected to Kieskompas (Krouwel, 2012). The propensities for each classification are summed to a total score for each participant.

After the propensity for each party, the respondents were asked which party they voted for during the last parliamentary election in 2012. Followed by the question what the chance is that they will vote for the same party as they did the last time. This was measured on a scale where ‘0’ means “not at all probable’ and ‘100’ means “very probable”.

As a manipulation check and to measure the SJ level, participants were asked to complete an eight-item questionnaire on the perceptions of the fairness, legitimacy and justifiability of the prevailing social system. This SJ scale is constructed by Kay & Jost (2003) (in the same study where they constructed the stereotype manipulation that is also used in experimental

condition A & B) in order to measure situational effects on SJ. By combining the eight items, a sum system justification score was calculated for each participant. The scores for the two reversed items in the SJ scale were reversed scored before the SJ score was calculated. Cronbach’s Alpha for this scale in this study is 0.73, meaning that the items are internally

(20)

19 related as a group. The results of the Cronbach’s Alpha together with results of a factor analysis on the items of the SJ scale can be found in appendix 2.

Filling in questions on descriptive characteristics finishes the experiment. The descriptive items ask for age, gender, highest finished degree, political left/right-orientation (0-100) and self-evaluated social status (0-100).

The Dutch texts of this survey including all items can be found in the appendix 3. Figure 1 shows a visual presentation of the routing of the different groups.

(21)

20

Results

The results section shows the results out of the data from the experimental survey in three sections. First, the characteristics of the survey sample and the effects of the control variables will be shown. Second, the differences between the manipulation groups and the control group will be shown and tested. After the differences between groups, data will be presented to explain certain differences. Third, the total data set as a whole without distinguishing groups will be analysed on the relation between SJ and conservative voting behaviour.

Characteristics of the sample

The online survey took place from 13 May 2016 until 21 May 2016. During this period 254 persons started the online survey, out of which 206 fully completed all the items. After removing three useless respondents, the remaining N for this study is 203. One respondent (respondent 18) was removed because all given scores were 0, even for age. Two respondents (respondent 56 & 128) were removed because of extreme scores (1’s or 100’s for nearly all items). From the remaining 203 respondents, 58% were men (118) and 42% were women (85). The mean age in this survey is 30,84 with a standard deviation of 12,59. The minimum age is 18 and the maximum 62. The largest part of the respondents has an age between 20 and 30 (±140). The mean age pulled up because of a solid group of ± 35 respondents with an age between 50 and 60.

The mean on the scale where the respondents place

themselves on the left/right-dimension is slightly more left wing with a mean score of 44,00 and a standard deviation of 23,32. Table 1 shows the frequencies and percentage of the respondents for their voting choice at the last parliamentary elections in 2012. Notable in this distribution are the high scores for ‘D66’ and

‘GroenLinks’ compared with the seats in the parliament and the polls at the time of the

survey. Also notable is the very low frequency of people that voted for the PVV in 2012 and

National Elections 2012 Propensities to vote

Frequency Percentage Mean Std. Deviation

VVD 39 19,2 36,42 33,733

Partij van de Arbeid 38 18,7 44,08 30,590 Partij voor de Vrijheid 2 1,0 9,24 19,74

D66 48 23,6 54,67 32,15 SP 13 6,4 30,81 28,59 CDA 8 3,9 26,74 27,53 ChristenUnie 1 ,5 12,47 17,24 GroenLinks 25 12,3 47,85 34,11 Other 2 1,0 I voted blank 5 2,5 I didn’t vote 22 10,8 Total 203 100,0

Table 1. Vote frequencies and mean propensities to vote in the sample

(22)

21 the propensity to vote for the PVV compared with current number of seats in parliament and the percentage in the polls at the time of the research.

On the self-evaluated social status scale, described as the social ladder (0-100), the mean score is 65,84 with a standard deviation of 15,75. The minimum score is 22 and the maximum 100. The distribution in respondent’s education level was: 29 Masters or PhDs. 100

Bachelors, 33 Intermediate Vocational Education and 41 secondary educations. Other options for this item that were not selected were: primary school, other or non. This makes the education level in the sample notable high.

The distribution of the random assignment to the different groups is shown in table 2. The high frequency for group E/5 is due to the randomisation, which at first divided the respondents equally to the two manipulation methods and the control group. After ±160 respondents, the

assignment to the different groups was corrected to only the manipulation groups, aiming to balance the frequencies.

Effects of control variables

On the control variables ‘gender’, ‘education level’, ‘left/right’, ‘social status’ and ‘age’, multiple regression analysis was carried out for the influence on the scores of the SJ scale and the variables of H1, H2 and H3. The results can be found in appendix 4. For the SJ score, significant positive predictors are found in education level (Beta=.140, p=.039 ), left/right-position (Beta=.295, p=.000 ) and social status (Beta=.249, p=.000 ). For the H1 variable score, a significant positive predictor is found in left/right-position (Beta=.634, p=.000 ). For the H2 score, significant positive predictors are found in education level (Beta=.191, p=.000 ), left/right-position (Beta=.277, p=.000 ), and social status (Beta=.162, p=.019 ). A

significant negative predictor for H2 is found in the age variable (Beta=-.172, p=.009). For the H3 variable score no significant predictors are found in the control multiple regression analysis.

Frequency Percent

Group A/1 (high SJ Manipulation) 41 20,2

Group B/2 (Low SJ Manipulation) 37 18,2

Group C/3 (High SJ Manipulation) 39 19,2

Group D/4 (Low SJ Manipulation) 31 15,3

Group E/5 Control group ₅₅ _27,1

Total 203 100,0

(23)

22

Differences between manipulation groups

Difference between manipulation groups in manipulation scores

After the participants read one of the four manipulation texts, they answered a few items corresponding with the text they read. The scores for these items enabled the option to determine if participants paid attention to the text. The scores on the SJ scale made it possible to check if the manipulation affected the SJ motives of the respondents.

Difference between group A & B (Complementary stereotypes manipulation)

In the complementary stereotypes manipulation condition, the survey included items to rate the person described in the vignette. In an analysis of covariance (ANOVA), to test the effect of condition on perceived contentment of the protagonist, the unhappy description was rated significant less content (M=32,36) compared with the poor but happy description (M=55,12) [F(1, 75)=20,23, p=.001].

To determine whether the complementary stereotypes manipulation affected the participants SJ scores, an independent sample t-test was conducted to compare the level of SJ in high (A) and low (B) SJ manipulation conditions. There was not a significant difference in the scores for the high SJ manipulation (M=441,36, SD=118,65) and the low SJ manipulation

(M=433,27, SD=140,87); t (76)=0,275, p=.78. These results suggest that the complementary stereotypes manipulation didn’t have much effect on the level of SJ of the respondent. The sum SJ score is minimum ‘0’ and maximum ‘800’. In the light of this range and the standard deviations, the mean sum scores for both groups are very close to each other.

Difference between group C & D (System Threat manipulation)

In the system threat manipulation condition, the survey included items to judge how content the persons in the article were with the political, economical and safety situation in their country. In an analysis of covariance (ANOVA), to test the effect of condition on perceived contentment with the political, economical and safety situation in their country, the system threat description was rated less content (respectively M=39,77, 39,92 & 39,48) compared with the system affirmation description (respectively M=64,71, 58,39 & 63,35). All three conditions differ significant (respectively F(1, 68)=15,33, p=.000; F(1, 68)=8,34 p=.005; F(1, 68)=13,21 p=.001)

To determine whether the system threat manipulation affected participants’ SJ scores, an independent sample t-test was conducted to compare the level of SJ in high (C) and low (D) SJ manipulation conditions. There was not a significant difference in the scores for the high (C) SJ manipulation (M=450,43, SD=122,10) and the low (D) SJ manipulation (M=443,64

(24)

23 SD=103,28); t (68)=0,247, p=0,81. These results suggest that the system threat manipulation didn’t have much effect on the level of SJ of the respondent. Also in this manipulation method, the mean sum scores for both groups are very close to each other.

Difference between manipulation groups in propensities to vote for conservative

parties (H1)

The manipulation of the respondents SJ motives was mead done to see if this affects the propensity to vote for conservative parties. The mean sum score of the propensities to vote for conservative parties enables it to see if these scores differ between the manipulation groups. The internal consistency of the conservative parties is checked in a factor analysis, which can be found in appendix 2. The Cronbach’s Alphas for the scores on the four parties was .637. This means that the scores for the conservative parties are internally consistent.

Difference between group A & B (Complementary stereotypes)

To determine whether the complementary stereotypes manipulation affected participants’ propensities to vote for conservative parties, an independent sample t-test was conducted to compare the sum score of propensities to vote for conservative parties in high (A) and low (B) SJ manipulation conditions. There was not a significant difference in the sum scores for the high (A) SJ manipulation (M=87,73, SD=69,79) and the low (B) SJ manipulation (M=72,84, SD=69,40); t (76)=0,94, p=0,35. These results suggest that the complementary stereotypes manipulation didn’t have much effect on the sum score for propensities to vote for

conservative parties.

Difference between group C & D (System Threat)

To determine whether the system threat manipulation affected participants’ propensities to vote for conservative parties, an independent sample t-test was conducted to compare the sum score of propensities to vote for conservative parties in high (C) and low (D) SJ manipulation conditions. There was not a significant difference in the scores for the high (C) SJ

manipulation (M=71,51, SD=65,43) and the low (D) SJ manipulation (M=75,74 SD=57,80); t (68)=-0,283, p=0,78. These results suggest that the system threat manipulation didn’t have much effect on the propensities to vote for conservative parties.

As shown in the descriptive characteristics of the sample, the PVV has divergent propensities scores. These divergent propensities may influence the potential difference between

manipulation groups. However, after excluding the PVV score’s in the sum score of propensities to vote for conservative parties, there was still no significant difference in both manipulation conditions. In the complementary stereotypes method, after excluding the PVV scores, there was no significant difference in scores for high (A) SJ manipulation (M=73,78

(25)

24 SD=60,08) and the low (B) SJ manipulation (M=68,19, SD=66,41); t (76)=0,39, p=0,70. In the system threat method, after excluding the PVV scores, there was also no significant difference in scores for high (C) SJ manipulation (M=65,41 SD=60,50) and the low (D) SJ manipulation (M=63,03, SD=49,28); t (68)=0.18, p=0,86.

Difference between manipulation groups in propensities to vote for government

parties (H2)

The manipulation of the respondents SJ motives was mead done to see if this affects the propensity to vote for government parties. The sum of the propensities to vote for government parties enables it to see if these scores differ between the manipulation groups.

To determine whether the complementary stereotypes manipulation affected the participants’ propensities to vote for government parties, an independent sample t-test was conducted to compare the sum score of propensities to vote for government parties in high (A) and low (B) SJ manipulation conditions. There was not a significant difference in the sum scores for the high SJ manipulation (M=77,39, SD=44,34) and the low SJ manipulation (M=70,97, SD=43,12); t (76)=0,65, p=0,52). These results suggest that the complementary stereotypes manipulation didn’t have much effect on the sum score of propensities to vote for government parties.

Difference between group C & D (System Threat)

To determine whether the system threat manipulation affected participants propensities to vote for government parties, an independent sample t-test was conducted to compare the sum score of propensities to vote for government parties in high (C) and low (D) SJ manipulation conditions. There was not a significant difference in the scores for the high (C) SJ

manipulation (M=74,98, SD=36,35) and the low (D) SJ manipulation (M=79,74 SD=39,80); t (68)=-0,521, p=0,60. These results suggest that the system threat manipulation didn’t have much effect on the propensities to vote for government parties.

Difference between manipulation groups in chance to make the same choice as

during the last election (H3)

The manipulation of the respondents’ SJ motives was mead done to see if this affects the chance to make the same choice as during the last election. The scores on the chance to make the same choice as during the last elections, enables to see if these scores differ between the manipulation groups.

(26)

25 chance to make the same choice as during the last election, an independent sample t-test was conducted to compare the mean chance score in high (A) and low (B) SJ manipulation conditions. There was not a significant difference in the mean scores for the high SJ manipulation (M=64,90, SD=29,23) and the low SJ manipulation (M=64,41, SD=29,84); t (76)=0,07, p=0,94. These results suggest that the complementary stereotypes manipulation didn’t have effect on the chance to make the same choice as during the last election. Difference between group C & D (System Threat)

To determine whether the system threat manipulation affected participants chances to make the same choice as during the last election, an independent sample t-test was conducted to compare the mean chance score in the high (C) and low (D) SJ manipulation conditions. There was not a significant difference in the scores for the high (C) SJ manipulation (M=62,26 SD=28,99) and the low (D) SJ manipulation (M=54,29 SD=30,02); t (68)=1,12 p=0,27. These results suggest that the system threat manipulation didn’t have much effect on chance to make the same choice as during the last elections.

Differences between manipulation and control groups

The experiment included one control group that didn’t get exposed to any manipulation. By including this control group next to the manipulation groups in a one-way ANOVA, it is tested if the scores of the manipulation groups differ from the control group. This is done for the depended variables: SJ Scale, voting propensities for conservative parties (H1), voting propensities for government parties (H2), and chance to make the same choice as during the last elections (H3). For variables of H1, H2 and H3 an independent sample t-test is done after the ANOVA, to look at the difference between the control group and the manipulation groups as a whole.

Differences between manipulation and control groups in sum SJ scores on the SJ

scale

A one-way between subjects ANOVA was conducted, to compare the effect of the group assignment on the sum score for the system justification motive in manipulation group A, B, C, D and control group E. There was not a significant effect of manipulation or control group on the sum score for the system justification motive for the five conditions [F(4, 198)=0.35, p=0.84]. These results suggest that manipulation group A (M=441,37 SD=118,65),

manipulation group B (M=433,27 SD=140,87), manipulation group C (M=450,44,

SD=122,10) manipulation group D (M=443,65, SD=103,28) and control group E (M=460,65, SD=101,12) didn’t differ much in mean sum scores on the SJ scale.

(27)

26

Difference between manipulation and control groups in sum score propensities

to vote for conservative parties (H1)

A one-way between subjects ANOVA was conducted, to compare the effect of the group assignment on the sum score of the propensities to vote for conservative parties in manipulation group A, B, C, D and control group E. There was not a significant effect of manipulation or control group on sum propensity scores to vote for conservative parties for the five conditions [F(4, 198)=1,99, p=0.098]. These results suggest that manipulation group A (M=87,73, SD=69,79), manipulation group B (M=72,84, SD=69,40), manipulation group C (M=71,51, SD=65,42) manipulation group D (M=75,74, SD=57,80) and control group E (M=105,47, SD=78,11) didn’t differ significant in mean sum scores on the SJ scale. However, the p value of .098 has to be noticed, which means that difference is near significant. Also the mean score for control group E is notable higher.

To determine whether this difference between the assignment to the manipulation or control group affected participants’ propensities to vote for conservative parties, an independent sample t-test was conducted to compare the mean sum score of propensities to vote for conservative parties for the manipulation groups and the control group condition. There was a significant difference in the sum scores for the control group condition (M=105,47,

SD=78,11) and the manipulation groups condition (M=77,22, SD=65,86); t (201)=2,579, p=0,01). These results suggest that the assignment to the control group or manipulation groups have significant effect on the sum score of propensities to vote for conservative parties.

Difference between manipulation and control groups in sum score propensities

to vote for government parties (H2)

A one-way between subjects ANOVA was conducted, to compare the effect of the group assignment on the sum score of the propensities to vote for government parties in

manipulation group A, B, C, D and control group E. There was not a significant effect of manipulation or control group on the sum propensity scores to vote for government parties for the five conditions [F(4, 198)=2,15, p=0.076]. These results suggest that manipulation group A (M=77,39 SD=44,34), manipulation group B (M=70,97, SD=43,12), manipulation group C (M=74,97, SD=36,54) manipulation group D (M=79,74, SD=39,80) and control group E (M=93,58 SD=40,08) didn’t differ significant in mean sum scores on the SJ scale. However, the p-value of .076 has to be noticed, which means that difference is near significant. Also the mean score for control group E is notable higher.

To determine whether the difference between group assignment in the manipulation or control group affected participants’ propensities to vote for government parties, an independent

(28)

27 sample t-test was conducted to compare the sum score of propensities to vote for government parties for the manipulation groups and the control group conditions. There was a significant difference in the sum scores for the control group condition (M=93,58 SD=40,08) and the manipulation groups condition (M=75,64, SD=40,85); t (201)=2,80, p=0,01). These results suggest that the assignment to the control group have significant effect on the sum score of propensities to vote for government parties.

Difference between manipulation and control groups in the score for chance to

make the same choice as during the last election (H3)

A one-way between subjects ANOVA was conducted, to compare the effect of the group assignment on the score for chance to make the same choice as during the last elections in manipulation group A, B, C, D and control group E. There was not a significant effect of manipulation or control group on the score for chance to make the same choice as during the last elections for the five conditions [F(4, 198)=0,974, p=0.42]. These results suggest that manipulation group A (M=64,90, SD=29,23), manipulation group B (M=64,41, SD=29,84), manipulation group C (M=62,26, SD=28,99) manipulation group D (M=54,29, SD=30,02) and control group E (M=56,82 SD=30,00) didn’t differ significant in scores on the SJ scale. To determine whether the difference between group assignment in the manipulation or control group affected participants’ chance to make the same choice as during the last election, an independent sample t-test was conducted to compare the chance scores for the manipulation groups and the control group conditions. There was not a significant difference in the sum scores for the control group condition (M=56,82, SD=29,99) and the manipulation groups condition (M=61,86, SD=29,46); t (201)=-1,08, p=0,28). These results suggest that the assignment to the control group or manipulation groups does not have a significant effect on the score for chance to make the same choice as during the last election.

Explanations for the high mean sum score in control group 5 on propensities to

vote for conservative and government parties

Control group 5 got notable higher mean scores for propensities to vote for conservative (H1) and government (H2) parties. Independent sample t-tests show that there is a significant difference between the control group and all manipulation groups together. To find an explanation for the differences, a closer look is taken in control group five on variance within the scores for H1 and H2 and the descriptive characteristics.

(29)

28 The scores in control group 5, for both H1 and H2 had no notable outliers or aberrant

distributions. At first sight, the descriptives of control group five don’t look very different from the manipulation groups. The main difference in the descriptives of control group five is that the D66 has ±10% more voters and the group of non-voters is nearly 4% bigger. These increases are at the expense of all the other parties, and are more or less equal with the percentage of the total sample within the a range of 5%.

For control group 5 the mean score on the left/right scale is with 46,04 slightly higher than the mean of the whole sample (M=44,00). By exploring the distribution of the left/right-scores in group five elaborately, it is notable that there is a large group of respondents on the right of the middle. However, the effect of this right-wing group on the mean was neutralized by a few extreme left-wing scores. The median in control group 5 is the highest compared with the other groups. The scores on the left/right scale are made visual in the boxplot in figure 2.

The closer look at the distribution of the left/right scores might explain the high mean sum scores for voting for conservative- and government parties. A correlation coefficient was computed to assess the relationship between the left/right scores and the sum score for propensities to vote for conservative parties. There was a positive correlation between the two Figure 2. Boxplot left/right scores for each group

(30)

29 variables, r=0.636, n=203, p=0.000. This means that there is a strong, positive correlation between the left/right scores and the sum scores for propensities to vote for conservative parties. The large group of respondents on the right of the middle on the left/right scale, in combination with the strong positive correlation, might explain the high mean scores for H1 in control group 5. However for H2, the correlation between the left/right scores and the sum score for government parties was significant but less strong (r=0.259, n=203, p=0.000). To see if the left/right scores are better predictors for the scores on H1 and H2 than the group assignment, a multiple regression analysis is executed. The result tables can be found in appendix 5. The left/right dimension is the strongest predictor but not the only one. All the independent variables (except the dummy for group 1 on H1 and the dummy for group 4 on H2) result as significant predictors for the scores on H1 and H2. This shows that the high scores for the control group 5 on H1 and H2 are not ‘only’ explained by the left/right characteristics of this group. Also the group assignment, and with this the manipulation treatment are significant predictors for the H1 and H2 scores (except group 1 for H1 and group 4 for H2).

Altogether, the results on the differences between groups show that there were no significant differences on the variables of H1, H2, H3 and the SJ scale between the high and low manipulation groups. By including the control group next to the manipulation groups within an ANOVA test, no significant difference between the groups was shown on the variables of H1, H2, H3 and the SJ scale. An independent sample t-test showed that there was a

significant difference between the control group and the manipulations as a whole on the variables of H1 and H2. A closer look at the characteristics of control group 5 showed that there is a large group of right-wing respondents, which might cause the differences. However, a multiple regression analyses shows that the high scores for the control group 5 on H1 and H2 are not ‘only’ explained by the left/right characteristics. Also the group assignments (except group 1 for H1 and group 4 for H2) are significant predictor for H1 and H2.

(31)

30

The relation between SJ and conservative voting behaviour (without

distinguishing groups)

For the total sample, without distinguishing the different manipulation and the control group, the data was analysed on the relation between SJ and conservative voting behaviour. For the three hypotheses the correlation and linear regression were analysed.

Correlation between sum SJ scale scores and sum scores for propensities to vote

for conservative parties (H1)

A Pearson product-moment correlation coefficient was computed to asses the relationship between the sum SJ scores and the sum scores for propensities to vote for conservative parties. There was a positive correlation between the two variables: r=0.301, n=203, p=0.000. The relation is significant positive, which makes that there is support found for H1. A

scatterplot summarizes the results (figure 3).

Figure 3. Scatterplot scores on the system justification scale and vote propensities for conservative parties (H1)

(32)

31 In the scatterplot two data points, high in the middle of the graph, seem out of line with the other scores and have a negative effect on the correlation. By excluding those two cases, the positive correlation between the sum SJ score and the sum score for propensities to vote for conservative parties becomes more strong: r=0.318, n=201, p=0.000 (R²=0.101).

To check if the correlation between SJ and voting propensities for conservative parties is not caused by a more general correlation between SJ and voting propensities, another analysis is done. For this analysis, the propensity scores for conservative parties are divided by the total sum score for all parties. The correlation between the sum SJ scores and the divided scores is less strong and not significant (r=0.095, n=203, p=0,178). This means that sum propensity scores for conservative parties are stronger and more significant related to SJ then the percentage of conservative propensity scores from the total amount of voting propensity scores.

Another check is done on how the maximum voting propensities on one of the conservative parties is correlated to the sum SJ score. This obviates the possibility that preference for conservative parties only comes forward in a high propensity for one of the conservative parties. This correlation between the maximum score for one of the conservative parties and the sum SJ scores is less strong (r=0.279, n=203, p=0.000), compared to the sum propensities for conservative parties.

(33)

32

Correlation between the sum SJ scale scores and the sum scores for propensities

to vote for government parties (H2)

A Pearson product-moment correlation coefficient was computed to assess the relationship between the sum SJ scores and the sum scores for propensities to vote for government parties. There was a positive correlation between the two variables, r=0.458, n=203, p=0.000. The relation is significant positive, which makes that there is support found for H2. A scatterplot summarizes the results (figure 4).

Figure 4. Scatterplot scores on the system justification scale and vote propensities for government parties (H2)

Also for this relation it is checked if the correlation between SJ and voting propensities for government parties is not caused by a more general correlation between SJ and voting

propensities. The propensity scores for government parties are divided by the total sum score for all parties. The correlation between the sum SJ scores and the divided scores is less strong: r=0.254, n=203, p=0.000.

What is the relation between System Justification and voting behaviour? : an experimental survey in The Netherlands