The economic crisis and political participation among young people

Transcription

1 The economic crisis and political participation among young people A multilevel analysis of political participation in 23 European democracies Ådne Hindenes Master thesis at the Department of Political Science Faculty of Social Sciences UNIVERSITY OF OSLO May 23, spring 2016 Word count: i

2 ii

3 The economic crisis and political participation among young people A multilevel analysis of political participation in 23 European democracies Ådne Hindenes iii

4 Copyright: Ådne Hindenes Year: 2016 The economic crisis and political participation among young people Ådne Hindenes Word count: Print: Copy Cat Forskningsparken, Oslo iv

5 v

6 Abstract This thesis studies two distinct dimensions of political participation, elite-directed and elite-challenging political participation among young people in the wake of the 2008 economic crisis. The two forms of political participation are based upon the classical concepts of Barnes, Kaase et al. (1979) conventional and unconventional political participation. The research question that is studied in this analysis is: Do changes in economic and social conditions affect the degree and type of political participation that young people engage in? I test two different theories, the civic voluntarism model and grievance theory. Both of the theories use resources to explain the political participation of individuals. The research question is answered through linear multilevel analysis of two rounds of the European Social Survey from 2008 and The analysis finds that there is an age gap in political participation, but that this age gap is between older individuals and the rest of the population. The main finding of the thesis is that resources are a prerequisite for both forms of political participation. This indicates support for the civic voluntarism model. vi

7 vii

8 Preface A master thesis is a milestone in 25 years of curiosity and search for knowledge. Many people have contributed to this search and deserve thanks for their contributions. First and foremost I would like to thank my thesis advisor Johannes Bergh. Without your good advice and our many interesting discussions the quality of this thesis would not have been nearly as good as it is today. I would also like to thank everyone at the Institute for Social Studies for providing a great and inspiring working environment. As a student I have been lucky to be able to take part in the academic community that is ISF. I would also like to thank the politics seminar at ISF for comments and advice on an early draft of this thesis. My sister, assistant Professor Hanne Fjelde, deserves credit for reading and commenting on a draft of the thesis. A saying goes, behind every man there is a strong woman, and I am no different. To my fiancée, Marianne, thanks for putting up with me for the last year. I am grateful for your emotional support and comments on my scholarly arguments throughout the years. Although there have been some ups and downs, you have been solid as a rock by my side. Without you I don t know if this thesis would have been completed. Some people have been there for a longer time and has provided support no matter what. Thanks to my mother Lisa, my father Per and my other siblings for providing an environment where curiosity could thrive and where knowledge about society and the world in general is valued. This thesis is a result of the support you have provided over the years. Throughout my academic journey at the University of Oslo, especially Karl Bjurstrøm, Erlend Langøren, Mads Motrøen and Morten Werring deserve special thanks. Without the discussions on problems in society, lunches and late nights at Blindern this author would have had a life filled with less fun and comfort. All errors and shortcomings are of course my own. Ådne Hindenes Oslo, 20th of May 2016 viii

9 ix

10 Contents 1 Introduction 1 2 Key concepts and theory The literature on political participation What is political participation? Who are young people? What is political trust? Economic insecurity Why European countries? Age and political participation Political trust and participation Economic conditions and political participation Summary Methodology and research design Data sources A note on index creation Dependent variables: Political participation Independent variables Age Political trust Economic insecurity Controls Quantitative analysis Why multilevel analysis? Assumptions of a multilevel model Treatment of missing data and summary statistics Analysis Bivariate relationships Are there age differences in political participation? Do differences in trust affect political participation? Does economic insecurity affect the political participation? x

11 4.2 Multilevel models Political participation among young people Summary of findings Implications and limitations Further research Bibliography 85 Appendices 93 List of Tables 3.1 Countries included in the analysis Correlation between different types of political participation Correlation between different types of political participation Factor analysis of different aspects of trust Factor analysis of different aspects of trust Summary statistics ESS Summary statistics ESS Mean values of different types of political participation by age (young) in Mean values of different types of political participation by age (old) in Mean values of different types of political participation by age (young) in Mean values of different types of political participation by age (old) in Bivariate correlation between political participation and trust Multilevel regression with Elite-directed participation as dependent variable, Multilevel regression with elite-directed as dependent variable, Multilevel regression with elite-challenging participation as dependent variable, Multilevel regression with elite-challenging as dependent variable, Findings A.1 Correlation between different aspects of trust A.2 Correlation between different aspects of trust A.3 Summary statistics of all variables A.4 Summary statistics xi

12 A.5 VIF-values for independent variables (2008) A.6 VIF-values for independent variables (2010) B.1 Mean values of different types of political participation by age within each country B.2 Correlations of aspects of trust and different types of political participation by country List of Figures 3.1 Formative and reflective indices Illustration of an hierarchal structure Scatterplot Elite-directed political participation vs. economic insecurity Scatterplot Elite-challenging political participation vs. economic insecurity Interaction effect between age (young) and trust in political authorities Interaction effect between age (young) and trust in political authorities Interaction effect between age (young) and change in youth unemployment rates and between age (young) and real GDP growth rate Interaction effect between age (young) and unemployment status xii

13 1 Introduction Over the last century a lot of attention has been given to political participation in comparative political science. During the last decades researchers have focused on the declining and stagnant electoral turnout. A broader perspective is taken in the literature on political action. This literature studies both conventional political participation like voter turnout, and unconventional political participation like protest participation (see Barnes, Kaase et. al. 1979). This thesis will study political participation from the latter perspective. Can factors connected to the recent economic crisis contribute to explaining differing levels of political participation among young people in Europe? The 2008 economic crisis brought with it economic insecurity and perhaps challenged the trust of the citizens in their governments. Few studies have studied political action in the wake of the 2008 economic crisis 1. A group that has been affected a lot by the crisis is young people. As far as this author knows, there are few or no publications that study political action in this group in the aftermath of the economic crisis. This thesis intends to fill this gap in the research literature. There are several trends that show that conventional forms of political participation has been in decline. Firstly, turnout in general elections is stagnant or in decline across established democracies (See Dalton 1996; Franklin 2004; Dalton 2014). Secondly, since the 1950 s party membership has been in decline across Europe. Fewer people are there to give a voice to the people when decisions are made in the parties. Lastly, party identification is also in a downward spiral (Dalton et al. 2011: 10). A group that is especially affected by these trends are young people. This age group shows the lowest levels in turnout, party membership and party identification (See Franklin 2004; Dalton et al. 2013; Milner 2010). The decline in conventional forms of political participation may indicate that young people are indifferent to this kind of politics. Yet, if we look beyond electoral turnout, it may be that young people participate in new and alternative ways. First, the youngest cohorts partakes more in web petitions through 1 Notable exceptions are Hooghe and Marien (2013) and Kern, Marien and Hooghe (2015). 1

14 social media sites, like Facebook, than older cohorts. Though some critics say that these petitions are not to be regarded as political activity per se as they concern topics that are not normally viewed as political (Milner 2010: 71). Second, we see a rise of protest movements across the globe, suggesting a rise in alternative forms of participation among young people There are several examples of this in Europe during the last five years, like the 15th of May-movement in Spain which protested against austerity measures. Another example is the London riots of 2011 where disenfranchised youths protested against the government by starting a riot. A final example is the anti-immigrant protests taking place in many German cities today. The rise in these alternative forms of participation may indicate that young people engage in different kinds of political activities instead of taking part in the conventional forms of political participation. The contrast between conventional and unconventional forms of political participation with respect to youths is interesting and warrants further study. Specifically, we do not know much about what may explain the shift from conventional to unconventional forms of participation. Adressing this gap, this thesis will study political participation among youths in Europe. More specifically it will explore the following research question: Do changes in economic and social conditions affect the degree and type of political participation that young people engage in? This question will be answered through a multilevel analysis of survey data from the European Social Survey (ESS). Data from 2008 and 2010 are used to compare participation before and after the economic crisis hits. The economic crisis was a shock which created worsening economic conditions, including higher youth unemployment. A multilevel approach is best suited taking into concideration the research question and the nature of the data. The hieratical structure is well suited for a multilevel analysis. The research question also points towards using multilevel analysis as there is an assumption that if there are any variations in individual participation these will reflect the aggregate level economic or social situations. What is political participation? Verba and Nie (1972) defines political participation as 2

15 any activity that has the intent or effect of influencing government action either directly by affecting the making or implementation of public policy or indirectly by influencing the selection of people who make those policies. Barnes, Kaase et al. (1979) defines unconventional political participation as elite-challenging activities such as demonstrations, petitions, boycotts, political strikes, occupations, street blockades and even violence against property and people in pursuit of political goals. Conventional forms of participation are elite-directed. These are the legitimate pathways of citizen involvement in politics, which are sanctioned and encouraged by the elites and by the rules of a democratic regime. Such as electoral participation, political meetings, campaigns, canvassing, et cetera. Dalton (2014: 54) shows that unconventional participation has become as widesperad as many forms of conventional forms of participation. Therefore I will use the terms elite-challenging and elite-directed forms of political participation. Both Inglehart (1977) and Barnes, Kaase et al. (1979: 524) predicted a decline in elite-directed participation and a corresponding rise in elite-challenging forms of participation. Both Inglehart (1997) and Norris (2002) show evidence for the hypothesis presented by Inglehart (1977) and Barnes, Kaase et al. (1979). Both their studies show that unconventional forms of political participation are replacing conventional forms of political participation among young people. From this I derive the first hypothesis that will be tested in this analysis: Young individuals are more engaged in elite-challenging forms of political participation than in elite-directed political participation when compared to adults. On the other hand crisis of democracy theory predicts that that there will be a gradual and widespread process of civic disengagement. This is due to an erosion of democracy with declining rates in all kinds of participation (Putnam 2001: 4). Milner (2010) refutes the claims of Inglehart (1997) and Norris (2002) on empirical grounds. The difference between those who are young and those who are not is far too small to justify the claim that unconventional participation is substituting conventional participation. This hypothesis will test if there are differences between young people and adults after the crisis. It is important to know if there 3

16 are differences in the aftermath in order for the research question to be of interest. In the aftermath of the crisis there were many protests against governments across Europe. This suggests that there is a level of distrust among the citizens. There are two main ways to view the relationship between trust and political participation. The first view assumes that citizens need political trust in order to participate in politics. Almond and Verba (1963) claim to show that a positive attitude towards the political system increases the likelihood of political participation, while a negative attitude leads to alienation from political activity. The second view states that distrust in the political system can serve as an incentive and motivator for elite-challenging political activity. Rosavallon (2008) claims that a critical outlook towards authority is necessary in liberal democracy and that this distrust in authority can motivate political participation. The basic premise of both these theories is that the way in which the respondents evaluate the political system will affect their willingness to act politically. This attiudional change is of importance when trying to explain variations in political participaton among young people. From this I derive a second hypothesis: Low levels of trust in the political system leads to high levels of elite-challenging political participation among young people. The economic crisis has brought along with it an environment of economical and political instability in many European countries. Summons of state leaders in Europe addressing the crisis have become commonplace during the years since the crisis erupted. Maslow (1943) pyramid of needs is a way to explain human behaviour from a psychological perspective. There are 5 levels in Maslow s pyramid and each constitutes human needs. Each of the needs at the foundation of the pyramid needs to be covered in order to care about the needs at the top. For example, a person needs physical and economic safety in order to care about securing self-realization. Inglehart (1977) uses Maslow s pyramid of needs in order to explain the rise of post-materialism and the spread of elite-challenging political participation across the globe. The idea of Inglehart (1977) is that in post-modern societies the populations basic security and economic needs are covered, so they can think about 4

17 self-actualization and take part in elite-challenging political activity. However, the political and economic situation in the aftermath of the 2008 recession might indicate the opposite of Inglehart s (1977) predictions. The predictions of Inglehart (1977) rest on the fact that elite-challenging political activity should rise in an environment with physical and economic stability, where the citizen s basic needs are covered. The economic crisis lead to job losses and many people experienced economic insecurity. This has also led to the rise of protest movements like the 15th of May-movement this as a response to economic insecurity. The assumptions of Inglehart (1977) can be tied to the civic voluntarism model of Verba et al. (1995). This theory assumes that participation is dependent on the personal resources of the individual, the attiudes that encourage participaiton and connections to groups who ask one to participate (Dalton 2014: 64). The contrast to the model is found in grievance theory which predicts that a lack of resources can function as an incentive for political participation (Gamson 1968). From this I derive a third hypothesis: Economic insecurity in European countries has brought with it elite-challenging political participation among young people. In the next chapter I will define and delimit key concepts and discuss theory related to political participation. In chapter three I will discuss the method and research design of this thesis. Chapter four will present the results from the empirical analysis and in chapter five I will discuss the implications of the results from the empirical analysis. 5

18 2 Key concepts and theory In this chapter I define and delimit key concepts related to the subject of this thesis. I will also present theory related to the substitution hypothesis, political participation and trust and political participation and the economy. However, I will first present a review of the literature on political action. 2.1 The literature on political participation The literature on political participation is broad and has a long history. Some of the literature is descriptive in nature, while other parts of the literature seeks to explain the variations in participation. 2 A debate within the literature has been if there is a trend upwards or downwards in turnout levels. Dalton (2014) shows a trend with declining and stagnant political participation across western democracies since the 1960 s. Blais (2007: 624) shows that there has been a mean downward trend in both established and new democracies from 1970 until Explanations of variations in political participation can be tied to different analytical levels. The micro level (individual explanations), the meso level (social and political context) and the macro level (institutional explanations) (Aardal 2001: 23-24). Studies have shown that variations in electoral participation can be tied to traits of the individual. These types of explanations often focus on the social and economic background of the individual (see Lipset 1960; Verba, Nie & Kim 1978). The trait of the individuals are resources that can prepare and motivate the individual for taking part in political activities. Dahl (1961) was the first to use the term political resources. According to him political resources are all types of traits that can be used in order to influence society (Strømnes 1993: 44-49). The civic voluntarism model which I discuss later on in this chapter is reliant on political resources. Wolfinger and Rosenstone (1980: 102) provide evidence to show that the level of education influences the level of political participation. The higher the level of education, the higher the probability of taking part in elections. This finding is confirmed in Dalton (2014: 65), who 2 For a broad review of the literature up until year 2000, see Aardal (2001) chapter 1. 6

19 argues that education taps personal resources that are strongly linked with participation. Self-interest among voters is used in rational choice theory to explain variations in electoral turnout. Downs (1957) claims that if the advantages for the individual voter outweighs the cost of voting, then the voter will take part in the election. If not, the voter will stay at home. The rational choice theory has motivated numerous studies of economic voting. In this research tradition voters either vote according to their pocketbook or they vote according to the national economic situation (Lewis-Beck and Stegmaier 2007: 519). A problem with models that use strict rational assumptions is that it is not rational to vote as the advantages selfdom outweigh the costs, yet a majority of people still vote. This is referred to as the paradox of voting. In order to try solving the paradox of voting some researchers have added some expressive aspects to the rational model. Riker & Ordeshook (1968) add civic duty to the model in order to explain why people still vote when it is not rational. Edlin, Gelman & Kaplan (2007) explains the same paradox by adding social preferences as the explanatory factor. With this logic it is rational to vote due to the good feeling the voters get by improving the well-being of others. Blais, Young & Lapp (2000: 183) claim that civic duty should be viewed as a competing explanation to rational choice. Voting can also be viewed as an expressive action. Teorell & Westholm (1999) are proponents of this view. They compare a model with strict rational assumptions to a model where voting is viewed as an expressive action and find that the expressive model is better at predicting turnout than the rational model. Other explanations focus on social and political contextual factors to explain electoral behaviour. These explanations have a longstanding history in political science. In Norway Rokkan & Valen (1962) are one of the classical contributors to this tradition. They provide evidence to show that geographical variations in electoral turnout and party membership can be tied to different social and economic properties and historical experiences. Campbell (1966) shows that the political frame of an election has a large impact on the turnout in said election. If there are small perceived differences between the political parties, the election will have a low intensity. Low intensity leads to low participation, while high 7

20 intensity leads to high participation. Both Blais et al. (2000) and Franklin (2004:162) put an emphasis on the character of the election. When the electorate feels that the election is a close one where they can have an impact on the result, they will participate. Fieldhouse et al. (2007) confirms that the closeness of the contest make a significant difference in turnout through the use of a multilevel analysis of data from the European Social Survey. In the same analysis they found that the overall turnout level of the country impacts the participation of young people. This can explain to us why voter turnout varies in between countries. The institutional context may also explain variations in political participation. Lijphart (2000: 318) connects the larges differences in political participation to the institutional framework, i.e. the different regulations for taking part in democracy. Fornos, Power and Garand (2004) find that the stronger the sanctions the greater the impact of compulsory voting legislation. However, they do not specify the effect of enforcement as such. The electoral system may also have an impact on the levels of participation. A traditional thesis is that systems with proportional representation will generate larger voter turnout than a majoritarian system (Lijphart 1999). This can be explained by the fact that people in a proportionate system are less estranged from the political system due to the fact that their vote will not be thrown away even though they vote for a small political party (Rach 2000:102). However, according to Blais and Aarts (2006) this view should be challenged. Having many parties in a party system (a feature of most PR-systems) decreases voter turnout as the probability of your vote having a decisive impact on the election remains extremely small. Yet, the literature on political participation does not only study voter behaviour. A broader perspective is taken in the literature on political action, which in addition to studying voter behaviour studies a wide range of different political activities like demonstrations, signing petitions and party membership. These types of studies focus on individuals or groups as a unit of observation rather than on protests or campaigns. This research tradition relates protest activity to socio-demographic, attitudinal and other characteristics of the respondents to investigate the predictors for participation in different forms of protest (Rucht 2007: 712) A seminal work in this body of literature is Barnes, Kaase et al. (1979) Political action: Mass 8

21 Participation in Five Western Democracies. They separate political participation into two distinct dimensions, unconventional political participation like taking part in a demonstration and conventional political participation like taking part in an election. The former is defined by being elite-challenging and the latter is defined by being elite-directed. Today the terms unconventional and conventional political participation seem a bit out of date as more and more people are taking part in so called unconventional political participation. A series of survey research has supported an increasing likelihood of using protest as a form of political participation (Jennings, van Deth et al. 1990; Norris 2002; Dalton 2002, 2004). Put in the words of Fuchs (1991) protest has become a normal part of politics. However, the terms elite-challenging and elite-directed political participation still make sense as one is controlled by the grassroots while the other is controlled by the elite. Having presented some of the classic perspectives on political participation, I define and delimit key concepts in this thesis and discuss theory related to elite directed and elite challenging political participation below. 2.2 What is political participation? In the research literature the definition of political participation has been discussed for a long time. According to Schumpeter (1942) and Sartori (1987) the role of citizens in the political process is limited to that of voting their leaders in and out of office. This is in line with what can be labelled as a liberal model of democracy. This definition of political participation is very thin and leaves very little to be studied empirically (Teorell 2006: 788). A broader definition, which I will use in this thesis was coined by Verba and Nie (1972). According to them political participation is: those activities by private citizens that (...) aim at influencing the government, either by affecting the choice of government personnel or by affecting the choices made by government personnel (Verba and Nie 1972). This is a definition, which is used widely throughout the literature on political participation. Although critics, like Bell (1974) and Huntington and Wilson (1976), have claimed that the 9

22 research on political participation was lacking a wider perspective with regard to the goal of democratization. The Verba and Nie (1972) definition adds an important aspect to the concept of political participation. This is the aim of influencing the government which opens up for a wide array of different political activities which have this as a goal. Participation in this definition is viewed as a instrumental act aimed at making the political system respond to their wishes (Teorell 2006: 789). According to Verba (1996) political participation is: a means by which governing officials are informed of the preferences and the needs of the public and are induced to respond to those preferences and needs. This can be called a responsive model of democracy. The most important contribution of Verba and Nie at the time was to add other varieties of citizen activities that could be conducive to responsiveness, in other words political participation beyond voting and campaign activity (Verba and Nie 1972: 2). However, they still kept to the notion that political participation was a indirect action towards the policy outcome (Teorell 2006: 789). What constitutes political activity? The boundary between political and non-political activity is not necessarily clear. Voluntary organizations that normally are not considered political can be an important linkage into political life. They can act as a platform for learning skills, organizational and communication, that are transferable to into politics. Or they can function as a recruitment platform for political organizations. The members of such non-political organizations are exposed to politics through these organizations (Verba, Schlozman and Brady 1995: 40). What separates these types of activities from political activity is, however, the aim. Political activity aims at influencing the government (Verba et al. 1995: 42). Many works in the field of political participation has a limited focus on what is considered political participation. Often the focus has been on the act of voting, as this is quite easy to measure through surveys and gives the possibility of checking the turnout against the actual voter register. In this thesis I look at political participation more broadly. An important distinction is between the classical concepts of conventional political participation and uncon- 10

23 ventional political participation. Barnes, Kaase et al. (1979) defines unconventional political participation as elite-challenging activities such as demonstrations, petitions, boycotts, political strikes, occupations, street blockades and even violence against property and people in pursuit of political goals. Conventional forms of participation are elite-directed. These are the legitimate pathways of citizen involvement in politics, which are sanctioned and encouraged by the elites and by the rules of a democratic regime. Such as electoral participation, political meetings, campaigns, canvassing, et cetera. A criticism of this definition is that what is viewed as conventional and unconventional political participation varies with time and place. What we today view as conventional participation in a liberal democracy, might not have been viewed this way 50 years ago (Verba et al. 1995: 47). That is the case with many of the types of activities that earlier was called unconventional political participation. An example of this is the use of petitions today compared with 50 years ago. Due to technological advances the use of petitions has grown as it is much easier than before to use this type of participation due to the internet. Petitions have now become a widespread form of political participation (Fuchs 1991) and hence it can no longer be called unconventional. Therefore this thesis will use the terms elite-directed activities and elite-challenging activities. The rise of the Internet also challenges the definition of Barnes, Kaase et. al. (1979). How should one characterize participation through the Internet? It can be viewed as elitechallenging political participation as it is a big challenge to the power of the elite. The Internet created a public space for a true deliberative democracy (Margolis 2007: 769). Blogs and other social medias clearly challenges the power of the media as a gatekeeper of opinions as any person today controls his or her printing press. Government officials have a hard time hiding their mistakes because the next day it could be posted online (Margolis and Moreno-Riaño 2007: 769). However, the Internet can also be viewed as elite-directed political participation. The elite also use the Internet and social medias as a way to communicate with and to organize their supporters. For example Barak Obamas first presidential campaign created MyBo as a platform to organize and energize their voluntaries (Dalton 2014: 58). This inspired parties accross the globe to create similar web platforms. Social media platforms, 11

24 like Twitter and Facebook is also used as means for party movements to communicate with the voters (Dalton 2014: 58). A third way of viewing the Internet and social medias is as a tool that enables both elite-directed political participation and elite-challenging political participation. This thesis has the latter view of the Internet and social medias as it enables both forms of political participation providing an easy way to organize and get messages through. 2.3 Who are young people? The main interest of this thesis is political participation among young people. In order to study this phenomenon the concept of young people has to be defined. This is not an easy task as the concept of young people does not have a clear cut definition over time. Due to changes in modern society it has become increasingly harder to define. Furlong (2013) defines young people as the socially constructed intermediary phase that stands between childhood and adulthood. It cannot be defined chronologically as a stage that can be tied to specific age ranges or to specific activities. Some tie it to the age of maturity. However, the age of maturity can differ from country to country. Others tie it to when people start to work, but this is becoming increasingly difficult. Because of increasing opportunities to get an higher education people tend to study longer before they start working (Furlong 2013: 3). Different governmental bodies have tried to define the concept through tying it to the age of a person. The Norwegian government defined it as the group between the ages of 12 to 26 (NOU 2011: 20). According to the United Nations, young people are defined as the group between the ages of fifteen to twenty-four (UN 1981). The African Union defined young people as the generational group fifteen to thirty-five (UN 2008). As among the governmental bodies there are different definitions in the research literature of who are to be considered young. Some researchers like Norris (2003: 11) define young people as the generational group from eighteen to twenty-nine. Others like Fieldhouse et al. (2007) define young people as the generational group from eighteen to twenty-four. The 12

25 people between the ages of 15 and 17 are excluded as they cannot take part in an election. In this thesis young people will be defined as Norris (2003) defines it. This because I test some of the theoretical assumptions of Norris (2003) on new data and because this definition will also give a larger N to draw inference from. The latter point is significant because young people are often underrepresented in surveys like the ESS. 2.4 What is political trust? Over the years there have been many different attempts at creating a definition of political trust, but no clear consensus has been reached. 3 The main disagreement in the debate on political trust is about how many levels or objects of trust there are (Bergh 2004: 370). David Easton (1965) separates trust into diffuse and specific trust. Easton (1965) views specific trust as satisfaction with the government s performance and policies, while diffuse trust is viewed as peoples satisfaction with the political system. He argues that there are three objects of trust: 1) the authorities, 2) the regime and 3) the political community (Easton 1968: 172). Gamson (1968) uses Easton s ideas to develop his own theory of Power and Discontent. According to Gamson (1968: 49-50) trust may be directed at four different levels: 1) the incumbent authorities, 2) the political institutions of a regime, 3) the public philosophy of a regime and 4) the political community. While Pippa Norris (1999) separates her concept of political trust into five different levels of political trust. Both Miller (1974) and Citrin (1974) focus on two levels of political trust: 1) the incumbent political authorities and 2) the political institutions. Bergh (2004) argues, with support from survey evidence, that in empirical research one should distinguish between two levels of trust: 1) trust in the political authorities and 2) trust in institutions. He argues that it is hard for the democratic citizens to distinguish fine masked concepts such as Norris (1999) political trust-concept. However, such additional levels of trust can be useful in theoretical discussions (Bergh 2004: 371). I find Bergh (2004) 3 For a broad review of the classical literature on political trust, see Bergh (2004). 13

26 argument to be compelling and therefore I will use his definition of political trust in my thesis. 2.5 Economic insecurity Economic insecurity is a term which has been frequently used in the aftermath of the 2008 financial crisis and the ensuing state debt crisis. All though it has been used by many authors and experts the concept has not been defined with precision. In the research literature there have been a multitude of definitions of what economic insecurity is, but no consensus has been reached (Bossert and D Ambrosio 2013). According to Osberg (2010: 3) economic insecurity is the anxiety produced by the lack of economic safety i.e. by an inability to obtain protection against subjectively significant potential economic losses. Jacobs (2007: 1) understands economic insecurity as the intersection between perceived and actual downside risk. While Stiglitz, Sen and Fitoussi (2009: 198) define economic insecurity as uncertainty about the material conditions that may prevail in the future. This insecurity may generate stress and anxiety in the people concerned, and make it harder for families to invest in education and housing. Bossert and D Ambrosio (2013) provides a sythesis of the above definitions in order to create a theoretical concept. They define economic insecurity as the anxiety produced by the possible exposure to adverse economic events and by the anticipation of the difficulty to recover from them (Bossert and D Ambrosio 2013: 1018). This definition involves past, present and future. The individual is insecure about the future, and this expectation produces anxiety in an individual. The resources we have today are important in determining what we can rely on in case of a future adverse event. The past plays a role in shaping the confidence in which we meet adverse events (Bossert and D Ambrosio 2013: 1018). 2.6 Why European countries? Other cases could have been used in this analysis. Inglehart (1997) use the World Values Survey that includes countries from the entire globe. However there are reasons why European 14

27 countries are chosen as cases in this analysis. Firstly, Europe has countries with high GDPs and highly educated populations. This gives a good posibility for comparasin. Secondly, many European countries were hit hard by the 2008 economic crisis and are still suffering due to its effects. This makes it possible to study the effects of change in economic insecurity on political participation among young people. As there are countries like Greece which were struck quite hard by the crisis and countries like Germany which got through the crisis quite well. Lastly, the data available through the European Social Survey is of high quality. 2.7 Age and political participation A prediction from Ronald Inglehart s seminal work The silent revolution : changing values and political styles among Western publics (1977) states that elite-directed forms of participation like electoral participation and party membership, are in decline while there is a corresponding rise of elite-challenging forms of participation such as participation in demonstrations and signing petitions. In their classic book on political action, Barnes, Kaase et al. (1979: 524) also predict this spread of unconventional forms of political participation and describe the trend through cross-sectional data. Inglehart (1997) provides evidence for the hypothesis through the use of longitudinal data from the World Values Survey. The book describes a trend with stagnant or falling electoral participation throughout the western world. The parties are losing their grip on the electorate and their membership rate is falling. The membership of political parties are about half of what it was only a couple of years ago (Inglehart 1997: 307). Parry et al. (1992: 234) in a study of political participation in Britain found that 40 percent of those drawn towards so-called direct action were under 29 years of age. This was the only form of participation where the youngest cohort was overrepresented. Earlier studies have shown a positive relationship between post-materialism and elitechallenging participation (Inglehart 1979: ). Inglehart (1997) show growing levels of mass participation. This is in stark contrast to the stagnant or falling levels of elite- 15

28 directed participation. The World Values Survey shows several trends that give insight into why the mass participation is on the rise. There is an increasing interest in politics, indicated by several measures. Rising numbers of respondents say that they are interested when asked directly and they say that they discuss politics. This paradox can be explained by the fact that there are two distinct processes going on: elite-directed participation and elite-challenging participation (Inglehart 1997: 308). The rise of the political parties in the 20th century can be tied to class and economy, while the trend we are seeing now can be tied to a shift in focus towards the issues that are linked to culture and life quality. In addition to this the younger generations are more postmaterialistic, thus they are less acceptant of the authority of the political parties (Inglehart 1997: 309). Post-materialist have a greater probability of taking part in elite-challenging activities than materialists. This is due to the fact that the post-materialist have grown up in a society without war and with economic safety. Therefore they view survival as inevitable, and they spend more time on activities that lead to self-fulfilment, like politics. Inglehart claims that this has led to a trend with growing levels of elite challenging political activity, especially among the younger generations (Inglehart 1997: 312). Norris (2002) confirms the findings of Inglehart (1997), elite challenging forms of political participation especially among the younger generations is becoming more widespread. Contrary to the popular belief that engagement in the population is dying, the trend is going towards a development, change and reinvention of engagement (Norris 2002: 4). The studies show that a growth in this type of participation is taking place. Elite-challenging participation is not only a trend of the 60s and 70s like some experts have claimed. Instead an increasing part of the population are taking part in elite-challenging activities (Norris 2002: ). Many forms of elitechallenging activities, like petitions, demonstrations and boycotts are quite widespread and have grown more popular over the last 20 years. This type of political participation is especially widespread among well educated in public sector in post-industrial societies (Norris 2002: 211). This is echoed in Quintelier s (2007) study of political participation in Europe and in Canada. She finds that young people are more likely to take part in activities like 16

29 demonstrations, petitions and boycotts. There are significant differences in the sample between young and old (Quintelier 2007: 176). Marien, Hoooghe and Quintelier (2010: 205) also find high levels of elite-challenging political participation among youths. 4 According to Inglehart (1997) and Norris (2003), democratic institutions have become less legitimate over time as a consequence of declining rates of elite-directed participation. They claim that the elite-directed participation is being replaced by elite-challenging participation. Putnam (1995) is one of the main proponents of the so-called crisis of democracy -school. This school claims that young people are more politically disengaged than previous generations and that if something is not done we can not hope for more engagement in this group. Young people tend to be generally less engaged in political activity and are on average less consistently involved in politics than earlier generations (Snell 2010: 260). A more nuanced view is found in Milner (2010). He does not think that young people are less engaged than previous generations. However, according to Milner (2010) it is very hard to find compelling evidence for the substitution hypothesis of Inglehart (1997). He claims that the substitution hypothesis is wrong both on empirical and methodological grounds. The difference in the data is too small to claim that a substitution between the two forms of political participation is taking place. 50 per cent of the respondents vote, while 28 per cent take part in elite-challenging activities. It is likely that some of these 28 per cent that are taking part in elite-challenging activities are also among the 50 per cent that vote. On the other hand elitechallenging political activity could still be replacing elite-directed political activities. As the 28 per cent that take part in elite-challenging activities could be growing and replacing the 50 per cent that vote. Empirically Milner (2010) backs up his claims with survey evidence from the International association for evaluation of educational achievement. 5 This survey shows that the respondents, with the exception of the US, have not had a change in values and that the youths in the Nordic countries have the least diversity in their participation. This 4 The analysis is run on data from the 2004-round of the ISSP survey. 5 The respondents of this survey are school children aged 14 years old in 28 countries (CIVED). 17

30 is contrary to the expectations that Inglehart presents. The value change is not happening and in the Nordics where the value change should be most present the types of participation are the least diversified (Milner 2010: 20). Milners (2010) view is challenged by more recent events. Across Europe and in the world in general there has been a rise in protest activities since the economic crisis. In May of 2011 an anti-austerity movement mobilized demonstrations across Spain. Young people were overrepresented in the organization and participation in these protests (The Guardian 2014). There have been almost daily protests against the austerity measures and the Greek politicians ever since the financial crisis of Also here young people were overrepresented. In the fall of 2011 there were riots in London and other major cities in the UK. These riots started due to the social conditions among the minority population in these cities. The protesters here were poor, young men of minority backgrounds. In the introduction I introduced a hypothesis about age and political participation. It is as follows: Young individuals are more engaged in elite-challenging forms of political participation than in elite-directed political participation when compared to adults. This is in line with Ingleharts (1997) and Norris (2002) views of the relationship between age and political participation. However, the financial crisis poses an interesting test for this hypothesis. As there is an expectation of a stronger effect as the crisis has had consequences for young people especially. 2.8 Political trust and participation The economic crisis has brought along protests against governments and this implies a level of distrust. Therefore the 2008 financial crises provide a good context to look at the relationship between trust and political participation. Trust has been put forward as an explanation for the levels of political participation by many authors. 6 There are two ways to view the 6 For a broad discussion on the subject see Hooghe and Marien (2013). 18

31 relationship between trust and political participation. One is to look at trust as a condition for all political participation. This suggest that trust in the political system is a condition for both elite-directed and elite-challenging political activities. Another way to view this relationship is to look at a lack of trust as a condition for individuals to abandon elitedirected activities and choose to take part in elite-challenging activities instead (Levi and Stroker 2000). The former perspective states that trust is needed in order to take part in political activities, no matter which form. Almond and Verba (1963: 27) claims that in order to become an active participant in politics the individual has to have a positive attitude towards the political system. Negative attitudes towards the political system leads to a rejection and this contributes to erode the legitimacy and the effectiveness of the democratic system. This leads to democratic instability over time (Almond and Verba 1963: 22). This is in line with a thinking that political participation requires resources. The latter perspective presents arguments that distrust in the political system can serve as an incentive and motivator for political participation. When the government and the political system as a whole are viewed as untrustworthy citizens may want to intervene through elitechallenging activity (Hooghe and Marien 2013: 133). This line of thinking was first brought forward in Barnes, Kaase et al. (1979) work on political action where they presented the distinction between conventional and unconventional political participation. Conventional participation is a manifestation of support for the political order (...) The presence of dissent in democratic societies is often indicated by unconventional modes of action (Barnes, Kaase et al. (1979: 444). There is an expectation that citizens with low levels of trust to a greater extent than others take part in elite-challenging activities that often are goal oriented. The causes are specific and are placed outside the political system (Klingemann and Fuchs 1995). Norris (1999) claims that the citizens more critical outlook towards the political system should not be viewed as a symptom of alienation from the political system. Rather it should be viewed as a symptom of the growth of a new generation with critical citizens. Rosavallon (2008) 19

32 takes this point even further claiming that a lack of trust is an important part of the liberal constitutional system. In a modern, liberal democracy it is not expected that the citizens should blindly trust political authority. It is however expected that they should be critical of the actions of the governments and politicians. A more extreme version of this presented by Hibbing and Theiss-Morse (2001:5) who claim that distrust is a incentive for all types of political participation. These arguments echoe the view of some of the classic literature on social movement participation, especially grievance theory, that focuses on the motivating role of distrust on political participation. Gamson (1968: 46-47) argues that a high level of trust in authorities implies lack of necessity of influencing them. According to Craig and Maggiotto (1981) there is no need to become active in politics without having major grievances towards the system. Hooghe and Marien (2013: 146) through use of survey data show that those citizens who trust and have knowledge of the political system are likely to take part in elite-directed forms of participation. This is in line with the theory of Almond and Verba (1963). In addition Hooghe and Marien (2013: 146) show that distrusting citizens are more likely to participate in a whole range of elite-challenging political activities. This lends supports the claim of Norris (1999) and Rosavallon (2008) that low levels of political trust does not imply an alienation from the political system, but rather indicate a trend towards different forms of interactions between the citizen and the political system. In chapter 1 I presented a hypothesis about the relationship discussed above. The hypothesis is as follows: Low levels of trust in the political system leads to high levels of elite-challenging political participation among young people. This hypothesis is in line with the perspective presented by Gamson (1968), Barnes and Kaase et al. (1979) and Norris (1999). These authors focus on the motivating effect of distrust and how distrust leads to a focus on different types of interactions with the political system. 20

33 2.9 Economic conditions and political participation The economic crisis brought about large changes in the economic conditions in many European countries. Hence it presents a good opportunity to look at the relationship between economic conditions and political participation among young people. It is not clear how economic conditions contribute to differing levels and types of political participation among young people. There are two main theories that predict what economic conditions do to political participation: grievance theory and the civic voluntarism model. 7 The grievance theory predicts that a lack of resources will have a motivating effect on the political participation of citizens (Gamson 1968). In the civic voluntarism model it is expected that the political participation of citizens is dependent on the resources available. A lack of resources gives a lower likelihood for political participation (Verba et al. 1995). Grievance theory gives valuable insight into the relationship between economic hardship and political participation. Gamson (1968) predicts that grievances are a strong incentive for collective action and for elite-challenging participation if the grievances are translated into political aims. According to the theory, poverty, economic deprivation and other adverse living conditions should stimulate protest activity. Klandermans et al. (2001: 42) defines grievances as feelings of dissatisfaction with important aspects of life such as living standard, income, employment etc. According to Klandermans (2001: 271) grievances are socially constructed. Collective action frames are sets of beliefs that serve to create a state of mind in which participation appears meaningful (Klandermans 1997: 17). The frames consist of three parts: injustice, identity, and agency (Gamson 1992). Klandermans (2001: 272) specify these frames: Injustice specifies a sense of outrage about the way authorities is treating a social problem. Identity specifies a common we, a group or a social category that feels 7 For a review of the literature see Dalton (2014). 21

34 deprived and angry, and a they, authorities that are to blame for the situation the group is in. Agency is the perceived opportunities, the belief that collective action can successfully change the situation. In addition to being a contextual predictor of protest activity, grievance theory can also be an explanation of individual protest behaviour. Protest is viewed as a response to societal problems and citizen dissatisfaction. The theory argues that when social conditions change it causes people to experience a relative deprivation that again causes the likelihood of protest and rebellion to increase significantly (Gurr 1968: 1104). The idea of grievance theory is that personal or collective dissatisfaction and grievances acts as an incentive for political participation in general and particularly elite-challenging participation. The contrast to the grievance theory is found in the civic voluntarism model. The basic premise of this model is that political participation is dependent on the individuals access to resources (Verba et al. 1995). In this model grievances are seen as ubiquitous and thus the key question is whether aggrieved groups have the resources to mobilize. Such resources include money, time, technical infrastructure, expertise and so on (Klandermans 2001: 272). McCarthy and Zald (1976) use organizational theory and concepts such as social movement sector and industry, and movement entrepreneurs in their attempt to theorize about resource mobilization by social movements. Their resource mobilization by movement organizations creates a vicious circle. In this circle one needs resources to maintain the structures that one needed to mobilize those resources. Buechler (1993) criticizes the resource mobilization model for having an instrumental outlook on protest behaviour. However, the idea that participation requires some resources has been followed up in the civic voluntarism model. In the model it is expected that a person who has resources, like time, money and civic skills to be active in politics, will use these resources to become more active in politics (Berinsky 22

35 2002). According to Verba et al. (1995: 271) citizens will become more likely to participate, but also more efficient when they do participate when the resources of money and time are brought together with civic skills. Both Inglehart (1997) and Norris (2002) use the theoretical assumptions of the civic voluntarism model in their research. The model can be related to Maslow s (1943) pyramid of needs where the most basic needs must be covered in order for an individual to care about self-actualisation. This means that needs related to security and economy must be covered in order to care about taking part in political activities. Kern et al. (2015) evaluate these theories through the use of survey data. Their results indicate that both theories might have some hold, though under different conditions. In times with good economic conditions it lends support to the civic voluntarism model. This suggests that in good economic times resources are necessary in order to participate. However, in times with economic shocks, like the economic crisis the study indicates that grievance theory might have some hold. This indicates that when the living conditions of the population worsen drastically there is a corresponding rise in elite-challenging activities. The last hypothesis presented in chapter 1 is tied to the above discussion. It states the following: Economic insecurity in European countries has brought with it elite-challenging political participation among young people. This is in line with the grievance perspective presented by Gamson (1965) and Gurr (1968) who claim that grievances, i.e. economic insecurity functions as an incentive for political action. It is also contrary to the view of Inglehart (1997) and Norris (2002) who both claim that resources are a necessity for political participation. 23

36 2.10 Summary In this chapter I presented the general literature on political participation. This was done in order to provide context to the discussion later in the chapter. I have defined and delimited four central terms: political participation, young people, political trust and economic insecurity. This was done in order to give a theoretical definition of the four terms. I discussed why study European countries. The relationship between age and political participation was discussed and tied this to hypothesis 1 which was presented in chapter 1. This to see if there is still an age difference in political participation. I discussed theory regarding trust and political participation and also tied this to my second hypothesis presented in chapter 1. This because trust or distrust has been put forward as a resource which can affect political participation among young people In the last section I discussed theory regarding the economy and political participation and tied this discussion to my third hypothesis presented in chapter 1. This hypothesis gives room to see if the economic insecurity affects the political participation of young people in Europe. Together the three hypothesis will provide an opertunity to study the research question. 24

37 3 Methodology and research design This chapter will describe the data, the methodological approach and the research design of this thesis. First I present the data used to conduct this analysis and discuss the reliability and validity of this data. Second I present the indicators I use in the analysis both on the individual level and on the country level. Finally I discuss why a multilevel statistical approach is the most effective and best way to provide answers to the research question. 3.1 Data sources The data used to address the research question is a combination of the European Social Survey (henceforth referred to as the ESS) and country level indicators collected from the World Bank. The ESS is a large cross-national survey of populations in European countries. It measures the attitudes, beliefs and behavioural patterns of the populations in the countries taking part in the ESS. The survey employs a random probability sampling. The survey is conducted through hour-long face-to-face interviews, which includes questions on a variety of topics repeated from the previous rounds of the survey. The universe of the survey is all persons 15 years of age and above that resides within a private household, regardless of nationality, citizenship, language and legal status (European Social Survey 2014a). In the analysis the respondents below the age of 18 are removed due to the fact that these respondents cannot take part in elections. The real GDP growth rate and the youth unemployment figures are obtained from the World Bank. This is due to large amounts of missing values on these variables in the ESS 2010-round. In order to secure that the same operationalization is used I obtained the 2008 variables from the World Bank as well. The number of countries that take part in the ESS differ between each round of the survey. In the 2008-round 31 countries took part in the survey. While in the 2010-round 27 countries took part in the ESS (ESS 2015). However, in the analysis 23 countries are included. The selection of the countries is done by the following two criteria. The Freedom House Index is used in order to exclude countries that are unfree. The second criterium is that the country 25

38 has to be included in both the and 2010-rounds of the ESS. 8 This contributes to insure that the results are comparable. In 2008 there were respondents in the 23 countries which are analysed. Due to missing values the number of respondents in the analysis is reduced to This is a reduction in the number of respondents in the analysis of 7,1 percent. In 2010 there were respondents in the 23 countries that are included in the analysis. Due to missing values the number of respondents in the analysis is reduced to This is a reduction in the number of respondents in the analysis of 7,5 percent. To assess the validity of the data one has to look at the sampling methodology. The ESS employs a random probability sampling from the population in European countries 15 years or older. Both the 2008 and rounds of the ESS has a high response rate for a survey and insures that the non-response rate is not a threat to the validity if the respondents do not systematically differ from the non-respondents (Couper and De Leeuw 2003: 165). This is hard to evaluate, but in order to ensure that a possible bias does not impact the results a design weight will be applied. The design weight is used in order to correct for bias in the sample design in the individual countries. In some countries certain groups have a greater chance of being picked and therefore the design weight is applied to correct for this selection probability bias (ESS 2014b). To assess the reliability of a data set that one has not taken part in developing is challenging, but there is some information which can shed light on the reliability of the data. Random measurement errors is a threat against the reliability of the data (Hellevik 2002: ). The large sample in the ESS should reduce the chance for random measurement errors affecting the results. Another threat to the reliability is the consistency of the field work and the founding of the survey (Hellevik 2002: ). The central committee of the ESS imposes strict and consistent methods of fieldwork and it is well founded (Kittilson 2007: 882). This should reduce the possible threat against the reliability. From this I conclude that the ESS has a high degree of reliability in the data. 8 Israel is included in both data sets and is rated as free by Freedom house, but is excluded as it is not a part of the European continent. 26

39 Table 3.1: Countries included in the analysis Country Freedom house index 2008 and 2010 Included Austria x Belgium x x x Switzerland x x x Cyprus x x x Czech Republic x x x Germany x x x Denmark x x x Estonia x x x Spain x x x Finland x x x France x x x United Kingdom x x x Greece x x x Croatia x x x Hungary x x x Ireland x x x Israel x x Lithuania Latvia x x Netherlands x x x Norway x x x Poland x x x Portugal x x x Romania x Russian Federation x Sweden x x x Slovenia x x x Slovakia x x x Turkey Ukraine x x 27

40 3.2 A note on index creation Figure 3.1: Formative and reflective indices Source: In this thesis I have to create indices in order to measure some key concepts. Indices can be reflective or formative in nature. In a reflective index causality flows from the latent construct to the indicator. This means that all items in such an index should be positively correlated. In a formative index causality flows in the opposite direction, from the indicator to the construct. This means that there are no assumptions about the patterns of intercorrelation between the items in the index (Coltman, Devinneym, Midgley and Venaik 2008: 1250). Both of these two types of indices are illustrated in the figure above. How does one decide when an index is reflective and when it is formative? There are different theoretical and empirical considerations that researchers have to make when creating indices. Coltman et al. (2008) suggest three theoretical considerations a researcher should make when creating an index. First, the nature of the construct has to be discussed. In the case of a reflective model the latent construct exists, while in the case of a formative model the construct is formed. Second, the direction of the causality between items and latent construct has to be considered. In the case of a reflective model the causality flows from the construct to the items, while in a formative model it is the opposite. Third, the characteristics of the items used to measure the construct. In a reflective model items are manifested by the construct, while in a formative model items define the construct. Researchers should also make empirical considerations. For example looking at the item intercorrelation is one empir- 28

41 ical consideration. Reflective models should have high positive intercorrelations between the items. While formative models can have any pattern of intercorrelation, but should possess the same directional relationship (Coltman et al. 2008: 1252). These considerations will be discussed in the next section when operationalization s of the indices are discussed. 3.3 Dependent variables: Political participation In chapter two of this thesis a theoretical definition of what political participation was presented. In this section I will operationalize this theoretical definition. First a short recap of the theoretical definition. Verba and Nie (1972) define all activities that directly or indirectly are aimed at affecting the political process as political participation. Barnes, Kaase et al. (1979) dichotomy of unconventional and conventional political participation is important in this thesis. Unconventional political activity is activities that are elite-challenging in nature. While conventional political participation is activities that are elite-directed in nature. As unconventional forms of participation have become more widespread this thesis will use the terms elite-directed and elite-challenging. Elite-directed participation is measured through an index of three variables which are related to party and electoral activity which are controlled by the elite. These are: partymembership, electoral participation and worked voluntary in a political party or political action group. Elite-challenging participation is measured through an index of variables connected to protest participation. The index consists of the following variables: boycott of products, demonstration and signing petitions. The variables in these indices are chosen due to theoretical considerations. In the case of elite-directed participation other variables in the ESS could have been added in addition to those that are already in the index. A variable such as wearing campaign buttons in the last twelve months could have been used. However, the other variables in this index require much more resources than wearing a campaign button and therefore I have chosen to exclude this variable. In the case of elite-challenging participation, there are no other variables in the ESS-data that can be used to operationalize 29

42 this concept. Both the index of elite-directed participation and the index of elite-challenging participation are formative indices. There are several reasons for this. In both instances the constructs (elite-directed and elite-challenging participation) are formed by the different items of political participation. The causality of the indices in both instances flows from the items of political participation to the constructs of elite-directed participation and elite-challenging participation. The different items of political participation contribute to define the constructs of elite-directed and elite-challenging participation. This means that the sum of the questions measure actual political participation. The intercorrelations are shown below and indicates weak intercorrelations. However, as stated previously in the case of formative indices the intercorrelation does not have to have a specific pattern. All of this points towards that the indices of political participation are formative in nature. Party membership, work in a party/action group, boycotts, demonstrations and signing petitions are derived from the same question in the ESS. The question is the following: There are different ways of trying to improve things in [country] or help prevent things from going wrong. During the last 12 months, have you done any of the following? Have you.... Electoral participation is derived from the following question in the ESS: Some people don t vote nowadays for one reason or another. Did you vote in the last [country] national election in [month/year]?. Both of the indices are on the individual level. The indices are both coded with values from 0-3, i.e. the number of activities the respondent has engaged in. Table 3.2: Correlation between different types of political participation 2008 Vote Work in party Member of party Petition Demonstration Boycott Vote Work in party Member of party Petition Demonstration Boycott Does the operationalization measure what I want to measure, elite-directed and elitechallenging political participation? Can measurement errors occur in the analysis? This can 30

43 Table 3.3: Correlation between different types of political participation 2010 Vote Work in party Member of party Petition Demonstration Boycott Vote Work in party Member of party Petition Demonstration Boycott be discussed through the concepts of reliability and validity. Reliability can be defined as the occurrence of random measurement errors. In other words how accurately the measurements are (Hellevik 2002: ). The questions asked about political participation in the ESS are closed questions with fixed alternatives. All respondents are given the same question and the same fixed alternatives. Therefore, there is no reason to expect that the survey should create anything else than minimal unsystematic variation between the respondents. One possible challenge to the reliability is the fact that all questions are coded by the interviewer. This could bring about some possible biases due to coding mistakes. The fact that the alternative answers to the questions are fixed contributes to reducing the risk of coding mistakes. Summing up, there is reason to expect that the operationalization of trust has a high degree of reliability. This means that the variation in political participation between the individuals in most of the cases are due to actual, substantial variations in political participation and not due to the characteristics in the collection of the data. Measurement validity (concept validity) is defined as the absence of systematic measurement errors and to what degree the operationalization captures different aspects of the theoretical concept (Adcock and Collier 2001: 531; Hellevik 2002: 51-52). One way that systematic measurement errors can occur is if the question asked is leading in nature. In my opinion the questions that are asked to measure the different aspects of political participation are neutral in nature and I cannot see how the phrasing of the question should lead to a bias in the respondents answer. Another way systematic measurement errors can occur is if the question only focuses on some aspects of the theoretical concept. The question asks about a broad range of different political activities. The elite-challenging political partic- 31

44 ipation variable could be a possible challenge to this as only three measures are included in the ESS which measures this concept. A third possible challenge to the measurement validity is the fact that participation is self-reported. This can lead to possible bias. As for example voter turnout can be over reported in a survey like this. However, this is a normal operationalization in the research literature also when using other data sets and it is hard to find the range of activities included her in any other data set. Examples of studies using these types of variables to operationalize politcal participation are Inglehart (1997), Norris (2002;2003), Hooghe and Marien (2013) and Kern et al. (2015). Therefore I concude that the operationalization of political participation has a high degree of measurement validity. 3.4 Independent variables In this section I will present the independent variables of the thesis both on the country level and the individual level and the controls for variables that have previously been shown to be associated with political participation of different types Age This is one of the most important explanatory variables in the thesis. The variable is on the individual level. In section 2.2 I presented a theoretical definition of what it means to be young. Furlong (2013) defines youth as the intermediary phase between childhood and adulthood. This is not a clear cut definition as it is hard to decide what this intermediate age is which we need in order to study this group. In this thesis young people are the people between years old. This is in line with the Norris (2003) definition. In this analysis age is coded into two different dummies, age (young) and age (old). This is done as the interest of this thesis is to see if there is a difference between young people and adults as defined in section 2.2. In age (young) the respondents between and 60+ are coded with the value zero and the young respondents aged between years are coded with the value one. In the other dummy, age (old), the young respondents between and the middle-aged group and over and are coded with the value zero and the middel-aged 32

45 are between years and are coded with the value 1. Both of these dummies are derived from the variable age of the respondent (calculated) which is included in both the 2008 and 2010-rounds of the ESS. This variable is calculated from the year of birth of the respondent. The question asked is the following: And in what year were you born?. Both the lack of a fixed alternative and the calculation of the age opens up for possible coding mistakes. This could challenge the reliability of the measurement. However, the likelihood of a coding mistake is considered to be small due to the rigorous methodological demands of the ESS. Despite the possible challenges to the reliability of the measure, I conclude that the age dummies have a high degree of reliability. The measurement validity of age (young) and age (old) is considered to be high. As they capture the concept and they present an opportunity to study differences in between the young people and adults. There is also little reason to believe that the question asked in the year of birth -variable to be leading in any way Political trust This explanatory variable is at the individual level. Bergh (2004) separates trust into two distinct levels: trust in political authorities and trust in institutions. Trust in political authorities is a political/economic evaluation of the governments, a party or politicians performance. While trust in political institutions is a trust towards the institutions of the political system as a whole. All variables in the two trust indices are based on the following question from the European Social Survey: Using this card please tell me on a score of 0-10 how much you personally trust each of the institutions I read out. The scale is from zero to ten, were zero is no trust at all and ten is complete trust. The scale is the same for the satisfaction with the democratic system, but here 0 is extremely dissatisfied and 10 is extremely satisfied (ESS 2014b). In the two tables below factor analysis of different aspects of trust in the 2008 and 2010 data is pretested. The loadings in the case of both the factor analysis indicates two factors. The loadings in factor one indicates that trust in parliament, trust in the legal system and 33

46 satisfaction with the democratic system together form a concept. Therefore trust in political institutions will be operationalized through an index that consists of: trust in the parliament, trust in the judicial system and satisfaction with the democratic system. All the variables in this index are related to institutions of a democracy and as such it will be called institutional trust. The strong loadings in factor two in both table 3.3 and table 3.4 indicates that trust in politicians and trust in political parties form a concept. Therefore, trust in the authorities will be operationalized through an index consisting of two variables: trust in political parties and trust in politicians. Table 3.4: Factor analysis of different aspects of trust 2008 Factor 1 Factor 2 Parliament Legal system Democratic system Politicians Parties Table 3.5: Factor analysis of different aspects of trust 2010 Factor 1 Factor 2 Parliament Legal system Democratic system Politicians Parties There are several facts that point toward the trust index as reflective in nature. First, trust is a latent construct that already exists. It does not need to be formed unlike the political participation constructs. Second, the causality in the model flows from the construct of trust to the items which asks about trust in different institutions and the political authorities. Third, the items are dependent on the construct in order to be understood. Fourth, the intercorrelations shows strong positive relationships between the items 9. This clearly indicates that the trust index is reflective in nature. All respondents in the ESS are given the same questions and alternatives. This means that 9 Two tables with correlations between different aspects of trust confirm this. These are included in the appendices. 34

47 the answers will vary in an unsystematic manner. The alternative answers to the questions included in both indices are fixed. As previously explained this reduces the risk of coding mistakes. Therefore I conclude that the two trust indices are of high reliability. In my opinion the questions asked about trust are not leading in nature and should not bring about biases in the respondents answer. A possible challenge to the validity of the indices is that there are limited questions asked about political trust in the survey and due to this fact it could be said that the indices only represent a small aspect of the concept of political trust. For example I could not find a good variable to measure the trust in government part of the theoretical definition of trust in political authorities. Despite this I would say that the two variables included in the index are good measurements of the concept as they give an indication of how the citizens evaluate the performance of the politicians and the politcal parties. I argue that the variables included in the trust in institutions index cover the concept they are supposed to operationalize. The variables included in both indecies are theoretically related to the concepts they are supposed to cover and due to the fact that they load strongly together in the factor analysis presented above. Therefore I conclude that the two indices of trust have a high degree of measurement validity Economic insecurity Real GDP growth rate is added as an independent variable. The variable is on the country level. In chapter two I presented a theoretical definition of economic insecurity. The real GDP growth rate is added as a part of the operationalization of economic insecurity. GDP (Gross domestic product) is an indicator of the economic activities of a country or a region. It reflects the total value of all goods and services produced less the value of goods and services used for immediate consumption in their production. It gives information about the size and performance of the economy. The real GDP growth rate is an indicator of the health of the economy. When the real GDP rate grows strongly, employment is likely increasing as companies hire more workers and people have more money to spend. When the real GDP rate is shrinking employment usually declines and people have less money to 35

48 spend. The real GDP growth rates are at market prices, percentage change on previous year GDP. It is adjusted for inflation in order to see if the economy actually has grown over the course of the year. The growth rates are presented in percentages (World bank 2016a). Real growth rate GDP is added from to the dataset from the World Bank due to a large amount of missing values in the variable already included in the ESS. The real GDP growth rate is obtained from the World Bank both in 2008 and in Therefore I expect that the calculations will be consistent in both datasets. The calculations also represent a threat towards the reliability as they are done by humans and this opens up the possibility of unsystematic measurement error. However, the World Bank has clear standards and methodology for creating these types of indicators. Therefore, I conclude that the measure of GDP is of high reliability. The real GDP growth rate is a part of measuring economic insecurity. It gives an indication of how the economy has developed from one year to another. It does not cover the entire concept of economic insecurity. However, together with the change in the youth unemployment rate and unemployment status it is a valid indicator for economic insecurity. Change in the youth unemployment rate is one of the main explanatory variables. This variable is on the country level. In section 2.6 I presented a theoretical definition of economic insecurity. Change in the youth unemployment rate is added in order to operationalize economic insecurity. It is expected that individuals will react to a change in the unemployment rate either with increased political participation or with This variable shows the difference between the youth unemployment rate the year the survey is conducted and the youth unemployment rate two years back for persons in the economically active population between ages This means that the change in the youth unemployment rate for 2008 is the difference between the youth unemployment rate in 2008 and in Employed persons are all persons aged 15 and over that during the reference week worked at least one hour for pay or profit, or who were absent from such work (World Bank 2016b). This variable is added on to the data set from the World Bank. General 36

49 unemployment could have been used as a indicator. However, youth unemployment is a better measure as the main interest of this thesis is political participation among young people. Unemployment status of the respondent is added as an independent variable. This variable is on the individual level. This variable is added to operationalize economic insecurity. The variable is coded as a dichotomous variable in the analysis. The question used to operationalize unemployment is the following: which of these descriptions applies to what you have been doing for the last 7 days?. There are seven alternative categories for main activities. Here unemployed (not looking for a job), unemployed (looking for a job) and permanently sick or disabled are coded as 1, while all other activities (employed, education, military service and housework) are coded as 0. This due to the fact that there is an expectation that the respondents that answer the activities that are coded as 1 will lack resources due to the employment relation they have Controls Gender is also added as a control as it has previously been found to have an effect on political participation. Burns et al. (2001) provides evidence to show that men are more prone to take part in elite-directed forms of participation than women. While Stolle and Hoogue (2011) shows that women are more prone to take part in elite-challenging forms of participation. Gender is a variable in the ESS which is coded zero (men) and one(women). Level of education is also added as earlier studies have shown that the level of education has an impact on the level of political participation. Wolfinger and Rosenstone in their classical work Who votes? from 1980 were one of the first to provide evidence of the positive relationship between the level of education and the level of political participation. There is an education variable in the ESS, this variable is coded in accordance with IAEAstandards (ESS 2014b). However, in the analysis it will be coded in the following way: primary education (1), secondary education (2) and tertiary education (3). 37

50 Jones-Correa and Leal (2001) show that religion plays an important role in determining the level of political participation. Therefore religion is added as a control variable. It is expected that having a religious belonging will lead to lower levels of elite-directed political participation and higher levels of elite-challenging political participation. Religion is a variable in the ESS and is derived from the following question: Have you ever considered yourself as belonging to any particular religion or denomination? It is coded in the following way: zero (no) and one (yes). 3.5 Quantitative analysis I use a quantitative approach to study political participation among young people in Europe. In a quantitative approach we abstract from particular instances to seek general description or to test causal hypotheses (King, Keohane and Verba 1994: 3). In this case the goal is the former of the two. A qualitative approach could have been used to study the phenomenon of political participation among youths. In a qualitative approach the goal would be to study a small number of observations to get an in depth understanding of a phenomenon (King et al. 1994: 4). An example of a good qualitative study of political participation among young people is Ødegård (2009) who studies political participation among young people in Norway. This illustrates the point of King et al. (1994: 5) that neither quantitative nor qualitative research is superior to the other. The differences between the two are stylistic and are methodologically and substantively unimportant (King et al. 1994: 4) I use a quantitative approach due to the fact that I wish to study political participation among young people in the whole of Europe. The goal here is generalizability across many countries. Due to the large N a quantitative approach is suited better for this purpose Why multilevel analysis? This thesis will use a multilevel statistical approach to answer the research question. It is possible to use other approaches, but there are several reasons why multilevel analysis is the 38

51 Figure 3.2: Illustration of an hierarchal structure Source: Created by the author. best approach in this instance. First the research question of the thesis suggests using multilevel analysis. The research question includes predictors on the individual and the country level. When dealing with this type of information multilevel modelling is the best way to handle the data. The multilevel model provides a coherent model that at the same time incorporates both individual- and group-level models. It is possible to use other models when trying to deal with this type of information. One way is to run a classical regression with predictors on both levels. A problem with this approach is that it does not correct for differences between countries beyond what is included in the predictors. An alternative to this is to apply a regression with country-dummies. However this would create severe problems of collinearity in the model (Gelman and Hill 2009: 8). Second the European Social Survey has an inherent multilevel structure. This means that the data is collected with a hierarchal structure, where individuals are clustered within countries. Statistical theory, both sampling-theory and Bayesian, says that inference should include the factors used in the design of data collection. This way of modelling data is a direct way to include indicators for clusters at all levels of a design. Another way to handle this type of data could be to apply least square or maximum likelihood regression. However, to apply either of these regression types would create problems of overfitting (Gelman and Hill 2009: 7). When a model is overfit it can cause regression coefficients, p-values and R- squared to be misleading. An overfit model is too complicated for the data available. The regression then becomes tailored to fit the specific sample, rather than reflecting the overall population (Frost 2015). 39

52 3.5.2 Assumptions of a multilevel model As with all statistical models, multilevel models, rely on a number of assumptions. The assumptions of a linear multilevel model build on the same assumptions as a ordinary least square model, but there are some differences as the model is more complex. The assumptions are: linearity, normality, homoscedasticity and independence of observations. Linear multilevel regression needs the relationship between the dependent variable and the independent variables to be linear. This means that the function forms have to be linear at each level. Another assumption when applying linear multilevel regression is that there has to be little or no multicollinearity. Multicollinearity happens when the independent variables are not independent from each other (Snijders and Bosker 2012: 153). As we can see from correlations between the different independent variables and from the squared VIF-values this is not a problem in model 1-3. However, in model 4 there are some issues with multicollinearity due to the inclusion of the interaction terms. 10 The variables affected by multicollinearity are age (young), the interaction between age (young)*trust in political authorities and the interaction between age (young)*trust in institutions. Some authors claim that this can be a problem due to an inflation in the standard errors and the p-values. One common, but misguided remedy for this is excluding the interaction terms which cause multicollinearity (Fox 1991: 15). Robert J. Friedrich (1982) defends the addition of interaction terms despite the risks of the standard errors and p-values being affected by multicollinearity. Bambor, Clark and Golder (2006) argues that much of the concern about multicollinearity in models including interaction terms has been overstated. Even if there is high multicollinearity the standard errors are still the true standard errors, the high multicollinearity simply implies a lack of information. They argue that the danger of inferential errors due to excluding constitutive terms from the model outweigh any possible benefits (Bambor et al. 2006: 70). Therefore I will include the interaction terms despite the threat of multicollinearity. Some of the assumptions are tied to the residuals. First, the level one residuals and the level two residuals have to be uncorrelated. This is a part of the independence assumption and is also referred to as 10 The correlation tables and the tables with the squared VIF-values are included in appendix A 2. 40

53 lack of autocorrelation. As this analysis uses cross-sectional data this assumption is already met. Second, in order to apply a linear multilevel regression there is an assumption of normality. This means that the residuals at level-one and the level-two random coefficients have to be normally distributed. Lastly, there is the homoscedasticity assumption. This means that the residuals at level-one has to have a constant variance and the level-two random coefficients has to have a constant covariance matrix (Snijders and Bosker 2012: 153). In order to test these assumptions I have run residual diagnostics as recommended by Joop Hox (2002: 22-27). All of the assumptions for the estimated multilevel models appear to be met to a sufficient degree. 3.6 Treatment of missing data and summary statistics Units with missing values only constitutes a problem in the analysis if the missing values are not randomly distributed (Little and Rubin 2002). As illustrated in the section on the data sources, there is not a lot of missing data in the ESS 2008 and 2010-rounds. However, the missing data can still pose a problem if it is not randomly distributed. The documentation from both rounds of the ESS indicates that there are no questions which led to a large amount of missing values. Therefore it is assumed that individuals with missing values are randomly distributed and they are therefore excluded from the analysis using listewise deletion. Table 3.6: Summary statistics ESS 2008 Variable Mean Std. Dev. Min. Max. N Elite-directed Elite-challenging Trust in institutions Trust in political authorities Gender Religion Education Age (young) Age (old) Unemployment status Real GDP growth rate Change in youth unemployment rate

54 Table 3.7: Summary statistics ESS 2010 Variable Mean Std. Dev. Min. Max. N Elite-directed Elite-challenging Trust in institutions Trust in political authorities Gender Religion Education Age(young) Age(old) Unemployment status Change in real GDP growth rate Change in youth unemployment rate In table 3.5 and 3.6 I present summary statistics for the and 2010-data. Elitedirected political participation has a high mean in both tables, but a lower variance. This is due to the inclusion of electoral participation in the index. Elite-challenging political participation, on the other hand has a low mean in both years, but a high variance. This is due to the fact that fewer people take part in elite-challenging political participation than in elite-directed political participaiton. As we can see from the tables the scale of change in the youth unemployment rate changes quite a lot between the two years, this is due to the impact of the financial crisis. The same degree of change is not observable in the real GDP growth rate variable. In this chapter I have discussed the methodology and the research design of this thesis. In the next chapter I will start the empirical analysis and testing of the hypothesis that have been presented. 42

55 4 Analysis In this chapter, I present the results of several types of analyses. As explained earlier the main form of analysis in this thesis are multilevel models. However, I start with some bivariate analyses. This in order to scrape at the surface and get a look at what the relationships between the two dependent variables and the different explanatory variables looks like. Before I start with the empirical analysis a short recap of the research question and the hypotheses is due. The research question of the thesis is: Do changes in economic and social conditions affect the degree and type of political participation that young people engage in? There are three hypothesis which will be tested in the analysis. These are: 1. Young individuals are more engaged in elite-challenging forms of political participation than in elite-directed political participation when compared to adults. 2. Low levels of trust in the political system leads to high levels of elite-challenging political participation among young people. 3. Economic insecurity in European countries has brought with it elite-challenging political participation among young people. 4.1 Bivariate relationships In this section I will do a preliminary bivariate analysis in order to get a glance at the form and the type of relationship between the two dependent variables and age, trust, youth unemployment and real GDP-growth rate. A bivariate analysis gives us a measure the association between two variables. This measure helps us understand the possible relationship between the two variables. There are different forms of bivariate analysis. One form of bivariate analysis is to check the central tendency (mean, median or mode) of a variable by another 43

56 variable. Another type of analysis is a correlation between two variables either in the form of a correlation table or in the form of a scatterplot. The scatterplot is a good way to graphically assess the relationship between two variables. While a correlation table gives us values to assess the strength and direction of the relationship Are there age differences in political participation? The relationship between age and political participation has as previously stated been a question of a long scholarly debate. The findings of Inglehart (1997) and Norris (2003) indicates a rise in elite-challenging political participation with a corresponding fall in elitedirected political participation. Below I present four tables with mean values of elite-directed participation and elite-challenging participation for two different age-dummies. The scale of both the elite-directed participation and elite-challenging participation variable is 0-3. Table 4.1: Mean values of different types of political participation by age (young) in 2008 Elite-directed Elite-challenging Table 4.2: Mean values of different types of political participation by age (old) in 2008 Elite-directed Elite-challenging Table 4.3: Mean values of different types of political participation by age (young) in 2010 Elite-directed Elite-challenging Table 4.4: Mean values of different types of political participation by age (old) in 2010 Elite-directed Elite-challenging The mean values presented above are quite interesting. They present a picture which is 44

57 a bit mixed. On the one hand they indicate an age gap in political participation as both the findings of Inglehart (1997) and Norris (2003) indicates. As is observable from tables 4.1 and 4.3 the differences in elite-directed political participation is quite large. The difference between those in the age and those over 30 is The difference in 2010 is still In table 4.2 and table 4.4 I use a different age-dummy, the people 60 and above has the high value. The age gap in participation that we saw in table 4.1 and 4.3 repeats itself in table 4.2 and in table 4.4. In both 2008 and 2010 the difference in elite-directed political participation between those between and those over 60 is 0.15 in favour of those over 60. This indicates that the people 60 and above to a larger extent take part in elite-directed political activities than those below 60. The gap in elite-challenging participation in table 4.1 and table 4.3 is a small one. The difference in the mean value is 0.03 in both 2008 and 2010 in the favour of those between 18 and 29. In table 4.2 and table 4.4 the difference in elite-challenging political participation is larger than in table 4.1 and table 4.3. In 2008 the difference between those between and those 60 and above is 0.2 in favour of those between In 2010 the gap is reduced to 0.16, still in favour of those between This indicates that there might be a age gap in participation. On the other hand, both types of political participation decreases from 2008 until This might indicate that the elite-challenging political participation is not on a continuous rise. However, the differences of the mean values from 2008 and 2010 are quite small and it is important to remember that they are of two different samples. The age-gap is still present and has not changed from 2008 until 2010 which indicates that the age gap is quite constant despite the fall in participation. Summing up, there is an age gap in elite-directed political participation. The young respondents participate less than the two older groups in both 2008 and There is also a small age gap in elite-challenging political participation in both 2008 and The age gap between the young and the two older groups in elite-challenging political participation is almost non-existing. However, the gap between the old and the other two groups in elite-challenging political participation is quite large. This indicates that the older respondents are less likely to take part in elite-challenging political activities. 45

58 An important point of this thesis is to look at differences between countries. Hence I have calculated mean values of different types of political participation by age within each country in table There is quite a lot of variation between the countries in the mean values of political participation by age. The following countries stand out with high mean values on elite-directed political participation both by age (young) and age (old): Cyprus, Denmark, 12 Sweden. Cyprus most probably stands out due to their voting rules. The interesting part is that there is still an age gap in Cyprus in elite-directed political participation. Both Sweden and Denmark have had constant high levels of turnout for decades (Pintor, Gratschew and Sulivan 2002: 79). The following countries stand out with low values on elite-directed political participation: Switzerland, Czech Republic, Estonia, Portugal. The following countries stand out with high values on elite-challenging political participation: Switzerland, Denmark, Finland, France, Norway and Sweden. Switzerland probably stands out due to its use of petitions in its direct democracy. However, there is something interesting in Switzerland. There is a reverse age gap when compared to the other countries. Here the older respondents to a larger extent than the young respondents take part in elite-challenging political activities. A possible explanation of this reverse age gap might be the way the direct democracy works. In Switzerland petitions are used in order to force referendums on certain laws 13. Elite-challenging participation (low values): Bulgaria, Cyprus, Estonia, Hungary, Poland, Portugal. Summing up hypothesis one receives initial support from this bivariate analysis. However, further multivariate analysis is needed in order to confirm the hypothesis. 11 Due to the size of table 5.6 it is presented in the appedices. 12 Their voting rules make taking part in elections mandatory. As such Cyprus gains high values on this variable (Gratschew 2004). 13 There are two forms of petitions that can force referendums in Swiss laws. An optional referendum requires signatures and is in oposition to a proposed law. A federal popular initiative requires signatures and can only be used to amend the constitution (The Swiss Authorities online 2016). 46

59 4.1.2 Do differences in trust affect political participation? In chapter two I discussed the theoretical relationship between trust and political participation. The discussion led to the formulation of hypothesis two. Below I present a first bivariate analysis of the relationship between trust and political participation. Table 4.5: Bivariate correlation between political participation and trust Elite-directed Elite-challenging Elite-directed Elite-challenging Trust in institutions.10**.10**.10**.10** Trust in political authorities.15**.10**.15**.08** Note: Significance levels ( p <.05 = *; p <.01 = **). The strongest correlation is the correlation between trust in political authorities and elitedirected political participation. This indicates that the respondents trust in political parties and politicians is a factor when choosing to vote, being member of and working for parties. The pearson s r for elite-directed political participation and trust in political authorities is.15 for both the 2008 and 2010 data. This relation has a moderate strength. The correlation between elite-directed participation and institutional trust is weaker. Here the pearson s r coefficient is.10 for 2008 and.10 for The direction of the relationships indicates that trust is a prerequisite for elite-directed political participation. It also indicates that trust in political authorities is more important than trust in institutions. This does not go against the expectations of hypothesis 2. The correlation between trust in political authorities and elite-challenging political participation is positive, but weak in both the 2008 and 2010 data. The pearson s r is.10 in 2008 and.08 in The correlation between trust in institutions and elite-challenging political participation is also positive, but weak in both the datasets. The pearson s r is.10 in 2008 and.10 in The positive direction of the relationship signals that trust is also a prerequisite for elite-challenging political participation. It also shows that trust in institutions is more important than trust in political authorities in explaining elite-challenging political participation. That trust is a prerequisite for elite-challenging political participation goes against the idea of hypothesis 2 that lack of trust is an incentive for elite-challenging political 47

60 activity. However, this is not enough to reject hypothesis 2 as it needs to be tested through an interaction which will be introduced in a multivariate model later on in the chapter. From table 5.7 we can see that the correlation between trust in institutions and elitedirected political participation is weaker than the correlation between trust in political authorities 14. There are some countries that stand out with stronger correlations. Cyprus has a significant, positive and moderate correlation between trust in institutions and elite-directed political participation with a pearson s r coefficient of.17 in 2008 and.13 in This means that elite-directed political participation to some extent is dependent on trust in institutions. When comparing the two years there is a weakening of the relationship between 2008 and This means that the trust in institutions is less important in explaining the elitedirected political participation in Cyprus in 2010 than in Estonia has a significant, positive and moderately strong relationship between trust in institutions and elite-directed political participation in both 2008 and in The pearson s r is the strongest in the 2008-data with.21, while it is a bit weaker in the 2010-data with.17. This indicates that trust in institutions explains more of the elite-directed political participation than in other countries. Here we also see a decline in the importance of trust in institutions as an explanatory factor between 2008 and In Croatia there is an increase in the importance of trust in institutions as an explanatory factor for elite-directed political participation. The strength of the pearson s r coefficient increases from.07 in 2008 to.20 in 2010, both coefficients are significant. This shows that trust in institutions explains more of the elite-directed political participation in 2010 than in In the Netherlands there is a positive moderate relationship. The strength is about the same in 2008 (.14) and in 2010 (.15). This means that the importance of trust in institutions as an explanatory factor remains stable between the two years. The correlation between trust in institutions and elite-directed political participation in Hungary is an interesting case. The correlation goes from being insignificant in 2008 to being strongly significant in However, the more interesting fact is that the 14 Due to the size of the table it is presented in the apendices. 15 As previously stated Cyprus has manditory voting and this could impact the results. 48

61 relation between the two variables changes direction. In 2008 there is a weak and negative relation between trust in institutions and elite-directed political participation. While in 2010 there is a moderate positive relation between the two variables. Although the relation in the 2008-data is insignificant this is quite interesting. It indicates that trust in institutions grows in importance as an explanatory variable for elite-directed political participation in Hungary. It also implies that something might have changed in Hungary between the two surveys. From trust in institutions having an insignificant, but very weak negative impact on elite-directed political participation in 2008 to trust in institutions motivating elite-directed political activities in There are no significant negative relationships between trust in institutions and elite-directed political participation. This suggests that there has been an overall increase in trust towards institutions which has a motivating effect on elite-directed political participation. The correlation between trust in political authorities and elite-directed political participation is the strongest in both years, but also her some countries stand out with stronger correlations than others. Bulgaria has a significant, positive and moderate correlation in both 2008 and The pearson s r is the same in both years with.21. This suggests that the importance of trust in political authorities in explaining elite-directed political participation remains stable between 2008 and 2010 in Bulgaria. This is also the strongest correlation of all the countries in both years. This means that trust in political authorities explains more of the elite-directed political participation in Bulgaria than in the other countries in both samples. Estonia has a significant, moderate and positive correlation between the two variables in both years. The correlation in 2008 is.21 and in 2010 the correlation is.18. The strength of the correlation decreases between 2008 and This indicates that trust in political authorities explains less in 2010 than in 2008 in Estonia. However, the correlation in Estonia is still stronger than the overall correlation in table 4.5. The Netherlands also stand out from the rest of the countries. The correlation between trust in political authorities and elite-directed political participation is significant, positive and moderate with a pearson s r coefficient of.18 in both years. This shows that the importance of trust in political author- 49

62 ities as an explanation of elite-directed political activities remains stable between 2008 and Norway also stands out with a correlation that is stronger than the overall correlation presented in table 4.5. The relation is significant, positive and moderate in 2008 with a pearson s r coefficient of.18. This is stronger than the overall correlation in that year. And suggests that trust in political authorities explains more of the elite-directed political activity in Norway than in Europe as a whole in However, the relationship weakens between 2008 and The relationship is still significant and positive, but the pearson s r coefficient is now.14. This implies that trust in political authorities explains less of the elitedirected political activity in Norway in 2010 than in 2008 and that it explains less of than in Europe as a whole. There are no countries with negative relationships between trust in political authorities and elite-directed political participation. Summing up, trust in political authorities has a motivating effect on elite-directed political activities and it explains more of elite-directed activity than trust in institutions. As explained previously the relationship between trust in institutions and elite-challenging political participation is significant and positive in Europe as a whole in both 2008 and in However, there are some countries that stand out from this trend. Greece is a country that was hit hard by the economic crisis. Here the relationship between trust in institutions and elite-challenging political participation is significant and negative in both 2008 and in In 2008 the pearson s r is This indicates that rising trust in institutions leads to lower elite-challenging political participation. This is in line with the grievance perspective where distrust is viewed as an incentive for elite-challenging political participation. In 2010 the pearson s r coefficient is This suggests that trust in institutions explains less of the elite-challenging political participation. However, it still gives support to the grievance perspective. Another country that was hit hard by the financial crisis is Spain. Also here the distrust has a motivating impact on elite-challenging political participation. The relationship is significant and negative in both 2008 and in Also here the pearson s r is weaker in 2010 than in 2008 which implies that trust in institutions explains less of the elite-challenging 16 See table 4.5 on page

63 activity in 2010 than in In Slovakia the same trend is present. There is a significant and negative relationship between trust in institutions and the relationship weakens between 2008 and Finland and France are examples of the opposite. Here the relationship is significant, negative and weak, but the strength of the relationship grows between 2008 and Ireland is an interesting case to look at. In 2008 the relationship between trust in institutions and elite-challenging political participation is significant, negative and weak. In 2008 distrust appears to explain some of the elite-challenging political participation. However, in 2010 the relationship is significant, positive and weak. This might indicate that there has been a shift in Ireland between the two years. From support of the motivating effect of distrust on elite-challenging political participation to support of the motivating effect of trust on elite-challenging political participation. But, it could also be that there is only a difference in the samples and not in the Irish public. Overall the data indicates that there is a difference in the relationship between trust in institutions and elite-challenging political participation between the countries that were hit hard by the economic crisis and those countries that were not hit that hard. The countries that were affected by the crisis the most give indications of support for the motivating effect of distrust in institutions on elite-challenging political participation. This finding goes against the finding in table 4.5 and is in line with the expectations of hypothesis 2. While overall the trend is support for the motivating effect of trust in institutions on elite-challenging political participation. The relationship between trust in political authorities and elite-challenging political participation is weaker overall than the relationship between trust in institutions and elitechallenging political participation. However, also this relationship is significant and positive for Europe as a whole. Also here there are some countries that stand out from the overall trend. In Slovakia there is a significant, negative and weak relationship between trust in political authorities and elite-challenging political participation in both 2008 and in This means that distrust in politicians and political parties can explain some of the elitechallenging political participation. In Ireland the relationship also here changes direction from a negative to a positive relationship between 2008 and In 2008 the relationship 51

64 is significant, negative and weak and suggests a support for the motivating effect of distrust on elite-challenging political participation. However, in 2010 the relationship is insignificant, very weak and positive. In Croatia there is a significant, negative and relationship between trust in political authorities and elite-challenging political participation in both 2008 and in This implies that distrust in parties and politicians to some extent can explain the level of political participation. In Finland there is a significant, negative and weak relationship in Both the direction and significance level changes from This indicates that there is a growing distrust in politicians in Finland and that this contributes to explaining the elite-challenging political participation here to some extent. Overall the comparison between the correlations in each country to some extent paints a picture where distrust in political authorities can motivate elite-challenging political participation. While in most countries trust in political authorities contribute to explaining the level of political participation Does economic insecurity affect the political participation? In this section I look at the relation between economic insecurity and different types of political participation. To access this relationship I have created four scatterplots of the bivariate relation between a aggregated version of the two political participation indices and real GDP growth rate. 52

65 Figure 4.1: Scatterplot Elite-directed political participation vs. economic insecurity 2008 In the graph on the upper left in figure 4.1 the relation between elite-directed political participation and real GDP growth rates from 2008 is evaluated. The fitted line indicates that there is almost no relationship between the two variables. The trend in the graph is confirmed by the insignificant correlation coefficient, the direction is negative. There are two clear outliers in the graph, Estonia and Cyprus. Estonia has low values on both the real GDP growth rate and elite-challenging political participation. Cyprus has high levels on both real GDP growth rate and on elite-directed political participation. The relationship between elite-directed political participation and change in the youth unemployment rate in 2008 is evaluated in the graph on the upper right in figure 4.1. The observations are spread out in the graph. From the fitted line we can observe a very weak positive relation between change in the youth unemployment rate and elite-directed political participation. However, the correlation coefficient indicates that the relationship is insignificant. Poland is an outlier here with a drop in youth unemployment of around 14 percent and a relatively low elite-directed political participation. The correlations between indicators of economic 53

66 insecurity on the country level and elite-directed political participation are very weak and insignificant in The direction of both the correlation between real GDP growth rate and elite-directed political participation and youth unemployment and elite-directed political participation is in line with what would be expected from grievance theory. However, as the correlations are insignificant this is probably due to coincidence. In the graph on the bottom left in figure 4.1 the relationship between elite-directed political participation and real GDP growth rates from 2010 is evaluated. The relationship appears to be negative, however the fitted line indicates that there is little to no relationship between the two variables. The correlation coefficient between elite-directed political participation and real GDP growth rates confirms this as it is insignificant and extremely weak. This is in line with the results from the same graph from There are two distinct outliers among the observations, Greece and Sweden. The financial situation of Greece explains their outlier status. However, Sweden has an unusually high value on both political participation and the real GDP growth rate. With the exception of Greece and Sweden the observations are quite clustered together in the plot. When we compare the plot from 2008 with the plot from 2010 we can see that the value of Greece, Ireland and Spain has shrunk. For example Greece has shrunk from.9 to.8 on the political participation scale. This might indicate less faith in elite-directed political participation in these countries. This would be logical as they were hit quite hard by the crisis. The graph on the bottom right in figure 4.1 looks at the relationship between elite-directed political participation and change in youth unemployment rates. The fitted line indicates a negative and weak relationship between the two variables. The correlation coefficient confirms the direction of the relationship, but it also indicates that the relationship is insignificant. The graph on the bottom right indicates that there is no relationship between elite-directed political participation and elite-directed political participation. This confirms the pattern from The correlation between elitedirected political participation and change in youth unemployment rate in 2010 is weak and negative. This indicates that low levels of youth unemployment will give high levels of elitedirected political activity. This indicates that you need resources in order to participate and 54

67 supports the civil voluntarism model. This would indicate that the citizens in the countries that are hardest hit with job loss are leaving the elite-directed forms of political participation as a reaction to the economic situation. Summing up the graphs in figure 4.1 indicate that economic insecurity on the country level has little impact on elite-directed forms of political activities. A possible explanation of this could be that the economic conditions on the country level are too abstract for the citizens to react to. Figure 4.2: Scatterplot Elite-challenging political participation vs. economic insecurity 2008 In figure 4.2 on the upper left the relationship between elite-challenging political participation and real GDP growth rates from 2008 are plotted out. The fitted line indicates a moderate and negative relationship between the two variables. The fitted line also indicates a linear relationship between the two variables. This is confirmed by the correlation coefficient which indicates that the relationship is also significant. There is a clear outlier in the plot. Estonia has a negative real GDP growth rate, but has one of the lowest rates of political participation as well. The negative real GDP rate should give a high value on elitechallenging political participation. On the upper right, in figure 4.2, the relationship between 55

68 elite-challenging political participation and change in youth unemployment is evaluated. The fitted line indicates a linear relationship. It also looks to be a moderate positive relationship between the two variables. The correlation coefficient between elite-challenging political participation and change in youth unemployment rate confirms that the relationship is moderate, positive and that it is significant. Poland, Sweden, Spain and Ireland are outliers. Poland has a large fall in the youth unemployment rate and a low level of elite-challenging political participation. The fall in the youth unemployment rate in Poland is almost 15 percent. Sweden stands out with a high value on elite-challenging political participation. While both Spain and Ireland stands out with a growth in the youth unemployment rate. The data indicates that economic insecurity on the country level might trigger elite-challenging political participation. From the graph on the upper left we can see that a rise in the real GDP growth rate gives a correspondent fall in the elite-challenging political participation. In the graph on the upper left we can see that a rise in the youth unemployment rate gives a corresponding rise in elite-challenging political participation. This indicates that a lack of resources can motivate people to take part in elite-challenging political activities. Both of these graphs are in line with what is the expectations from grievance theory. It also gives support to the expectations in hypothesis 3. This is an interesting finding as it is not what is to be expected if one views the civic voluntarism model and grievance theory as theories dependent on the economic circumstance. In figure 4.2, the scatterplot on the bottom left looks at the relationship between elitechallenging political participation and the real GDP growth rate in From the fitted line we can observe that the relationship between elite-challenging political participation and the real GDP growth rate appears to be negative and moderate. This is confirmed by the correlation coefficient which also informs us that the relationship is significant. The fitted line also indicates that the relationship is linear. There are some cases with values that stand out. Greece has a clear fall in the real GDP growth rate. With a fall of over 5 percent it stands out, but the value on the elite-challenging political participation is higher than expected when the relationship is positive. Sweden stands out with both a high value on the 56

69 elite-challenging political participation and with a high real GDP growth rate. Poland and Slovakia have high real GDP growth rates, but they have quite low rates of elite-challenging political participation. Most of the countries are clustered around 0 percent on the real GDP growth rate. The scatterplot on the bottom right evaluates the relationship between elite-challenging political participation and the change in youth unemployment rates from The fitted line indicates that there is a moderate and negative relationship between the two variables. The correlation coefficient has is significant. This confirms what the fitted line indicated. The fitted line also shows that the relationship is linear. Spain is an outlier with high values on elite-challenging political participation and a strong growth in the youth unemployment rate. This is interesting as this is what would be expected if economic insecurity fuelled elite-challenging political participation like in The correlations have changed direction since 2008 and they have grown in strength as well. In 2010 a low real GDP growth rate indicates a low value on elite-challenging political participation. While a growth in the youth unemployment rate gives lower elite-challenging political participation. This indicates that rising economic insecurity has negative impact on elite-challenging political participation in It supports the idea that citizens need resources in order to participate which is in line with the civic voluntarism model. This challenges the idea in hypothesis 3 that economic insecurity gives rising levels of elite-challenging political participation among young people. However, the graphs are not a test of hypothesis 3 as this requires a interaction between the two economic insecurity variables with age (young) to test. The general trend that we can observe from the scatterplots is that the correlation between elite-challenging political participation and the two economy variables is stronger than the correlation between elite-directed political participation and the two economy variables. The relationship between elite-directed political participation and economic insecurity on the country level is generally weak or non-existing. While the relationship between elitechallenging political participation and the two economic insecurity variables is generally moderate. However, the scatterplots gives an unclear picture of what the relationship between elite-challenging political participation is. On the one hand, the 2008-data I find support for 57

70 a grievance relationship between the elite-challenging political participation and economic insecurity. This lends support to hypothesis 3 and suggests that the citizens turn to elitechallenging political participation when in economic desperation. On the other hand, in 2010 I find support for the civic voluntarism model. As the relationship between elite-challenging political participation and the real GDP growth rate is positive. While the relationship between elite-challenging political participation and change in the youth unemployment rate is negative. This challenges the idea in hypothesis 3 and suggests that the citizens are dependent on resources in order to take part in elite-challenging political participation. As previously stated this does not mean that hypothesis 3 can be thrown away as it needs to be tested through an interaction. I have now looked at the different bivariate relationships. This is done in order to look at the relations between the dependent variables and the different independent variables. In the next section I develop the multilevel models which will be used to study the different hypotheses that were presented earlier. 58

71 4.2 Multilevel models Below are four tables with multilevel regressions from two years, 2008 and 2010, with two different dependent variables, elite-directed political participation and elite-challenging political participation. The baseline model (model 0), this means that the model is empty with the exception of the dependent variable. This model is created in order to get a model to compare the results with. Model 1 consists of the dependent variable, age(young) and age(old) which are two of the main explanatory variables. In addition gender, education and religion are added as control variables. These variables are added in order to control for variables which have been shown to affect political participation in earlier studies 17. Age(young) and age(old) are dummy variables with year olds and 60+ as the high values. Model 2 consists of a dependent variable, age(young), age(old), unemployment, trust in political authorities and trust in institutions. The two trust variables added in model 2 are different aspects of trust and are of importance when I am going to evaluate hypothesis 2. While unemployment on the individual level is important when looking at hypothesis 3. Model 3 includes real GDP growth rate and change in youth unemployment in addition to the aforementioned variables. I include these in order to look at economic performance variables on the country level, these are of importance when we assess hypothesis 3. Model 4 includes interactions between age(young) and the other independent variables in conjugation with the aforementioned variables. The interactions between age (young) and trust in institutions, and age (young) and trust in political authorities are added to evaluate hypothesis 2. The two interactions between age (young) and real GDP growth rate, and age (young) and change in youth unemployment rates are added to look at how country level indicators of economic insecurity among young people affect the political participation. While the interaction between age (young) and unemployment status is added in order to evaluate the effect of a individual level indicator of economic insecurity among young people on political participation. All of the last three interactions are added in order to test hypothesis See the discussion in section for a justification of why these variables are controlled for. 18 I have run multilevel binomic logistic regresions with alternative dependent variables to check the robustness of the results. Overall the results of the logistic regresions were in line with the findings from the linear multilevel regresions. 59

72 Table 4.6: Multilevel regression with Elite-directed participation as dependent variable, 2008 (0) (1) (2) (3) (4) Elite-directed Elite-directed Elite-directed Elite-directed Elite-directed Age(Young) (.021) (.022) (.022) (.043) Age(old) (.014) (.014) (.0136) (.014) Gender (.009) (.008) (.008) (.008) Education (.007) (.007) (.007) (.007) Religion (.017) (.017) (.017) (.016) Unemployment (.010) (.010) (.012) Trust in political authorities (.003) (.003) (.003) Trust in institutions (.004) (.004) (.004) Real GDP growt rate (.020) (.021) Change youth unemployment (.010) (.011) age(young)_trustpa (.007) age(young)_trustinst.004 (.006) age(young)_unemp.010 (.020) age(young)_unra.001 (.009) age(young)_grgdp.004 (.015) Constant (.026) (.037) (.046) (.051) (.053) SD(Constant) (.018) (.023) (.022) (.021) (.020) SD(Residual) (.012) (.011) (.011) (.011) (.011) Observations Groups AIC BIC ICC.042 Robust standard errors in parentheses p < 0.1, p < 0.05, p <

73 Table 4.6 presents results from a multilevel analysis of the 2008 round of the ESS with elite-directed political participation as the dependent variable. The interclass correlation (ICC) value is.042 and this indicates that the country level contributes 4,2 per cent of the variance. This is quite low, but there is no clear rule that says how high the value of the ICC has to be in order to use multilevel analysis. The low ICC-value is in line with what we saw in the bivariate graphs between elite-directed political participation and country level variables connected to economic insecurity. The country level explains very little of the elitedirected political participation. Model 1 includes age(young), age(old), gender, education and religion into the model. The direction of gender is as expected for elite-directed political participation, men participate more than women. Education has a positive impact on elitedirected participation which is in line with the findings of Wolfinger and Rosenstone (1980). Having a religious belonging leads to lower levels of elite-directed political participation. From the coefficient of age (young) we can observe that citizens in the age group between years old to a lesser extent than the age groups of years and 60+ take part in elitedirected political participation. This is in line with the previous research in the field which found that young people are less involved in elite-directed political participation than adults (Norris 2003). This finding indicates that there is a age gap between young people and adults. Where adults are more engaged in elite-directed forms of political participation than young people. The coefficient of age (old) indicates that the group that is 60 years and older take part in elite-directed forms of political participation to a significantly larger extent than the two groups below the age of 60. This indicates that there is also a age gap between the group that is years old and the group that is 60 years and older. The likelihood of taking part in elite-directed political participation increases with age. Simply put, the older you get the more you participate in elite-directed forms of political participation. This finding is in line with what Norris (2003) found in her study. It also gives some initial support for hypothesis 1. However, this model alone does not test hypothesis 1 as this requires a comparison between elite-directed political participation and elite-challenging political participation. It is not clear what type of trend this represents. Since this study does not look at longitudinal data, 61

74 it is not possible to say if this is a result of a generational trend or if it is related to the life cycle of the respondent. In model 2 unemployment, trust in political authorities and trust in institutions are added to the model. It now consists of these variables in addition to the variables included in the previous model. These are added to the model to see if individual grievances and the degree of trust affects the citizens participation in elite-directed political activities. A units increase in trust in political authorities leads to a increase in elite-directed political participation, all other variables held constant. Trust in political authorities appears to have a motivating effect on elite-directed political participation. This indicates that trust in politicians and political parties is needed in order to participate in elite-directed forms of political activities. This is not unexpected as the people who do have trust in the political elites also should be expected to have propensities towards more elite-directed forms of political participation. The direction of the relationship between trust in institutions and elite-directed forms of political activities appears to be negative. This would indicate that if a citizen has trust in institutions this would lead to a rejection of elite-directed forms of political participation. However, the relationship between the two variables is insignificant. If it were significant this finding would be interesting as it would suggest that growing trust in societal institutions would lead to a decrease in elite-directed political participation. That trust in political authorities is required for taking part in elite-directed forms of political participation is in line with the findings of Almond and Verba (1963: 27). It also challenges the position of Hibbing and Theiss-Morse (2001). They claim that distrust is a motivating factor for both elite-directed and elite-challenging forms of political participation. This is quite interesting as it indicates support for a civic voluntarism perspective of trust when it comes to elite-directed political participation. This finding does not go against hypothesis 2 as this hypothesis is concerned with elite-challenging political participation. The results indicates that the unemployed take part in elite-directed political participation to a lesser extent than other groups. On the one hand this appears to be in line with what one would expect to see if the civic voluntarism model explained political participation. According to Verba et. al. (1995) a individual needs 62

75 resources to participate in political activities. On the other hand Gurr (1968) argues that when social conditions change that causes people to experience a relative deprivation that causes the likelihood of protest and rebellion to increase. If this is the case one would not expect it to motivate elite directed political participation. However, it is expected that this would motivate elite challenging political participation. This will be looked at below. In model 3, two country level variables, real GDP growth rate and change in youth unemployment, are added to the regression. None of these variables are significant when added to the regression. Despite this the direction of these two variables is interesting. An increase in real GDP growth rate will give a increase in elite-directed political participation. This indicates that growth in the economy will be associated with a growth in elite-directed political participation. An increase in the youth unemployment rate is associated with growth in elite-directed political participation. The individual level variables still have the same values and the same significance level and direction. In model 4, I add five different interaction terms to the regression in addition to the variables added in the previous models. Out of the five interaction terms only one is significant. This is the interaction between age (young) and trust in political authorities. Bambor et al. (2006) suggests using margin plots to evaluate interaction terms as the coefficient in a regression only gives a snapshot of what the relationship looks like, while a margin plot can show what the relationship looks like for a range of values. Therefore I have created a margins plot that illustrates the interaction term between age (young) and trust in political authorities. 63

76 Figure 4.3: Interaction effect between age (young) and trust in political authorities Figure 4.3 serves to illustrate that the age-effects are conditioned by the trust in political authorities in We can clearly see that the marginal effect of age increases as we move along the horizontal axis, from low levels of trust in political authorities to high levels of trust in political authorities. The interaction indicates that there is a difference in elite-directed political particpation by age for different levels of trust. Trust matters more for the political participation of adults than for young people. As we can see in the graph from the increasing difference in political participation. For low levels of trust there are much smaller differences in political participation between the two groups than for high levels of trust. This is in line with the expectations of Almond and Verba (1963) as they expect trust to be a motivator for elite-directed political participation. This is the expected type of relationship between trust in politcial authorities and elite-directed political participation. It also illustrates the age gap in political participation. 19 The figure shows marginal effects controlled for the other variables in the regression. 64

77 Table 4.7: Multilevel regression with elite-directed as dependent variable, 2010 (0) (1) (2) (3) (4) Elite-directed Elite-directed Elite-directed Elite-directed Elite-directed Age(young) (.021) (.021) (.021) (.057) Age(old) (.012) (.011) (.011) (.011) Gender (.009) (.009) (.009) (.008) Education (.007) (.007) (.007) (.007) Religion (.013) (.012) (.012) (.012) Unemployment (.015) (.015) (.015) Trust in political authorities (.002) (.002) (.002) Trust in institutions (.004) (.004) (.004) Real GDP growth rate (.008) (.007) Change youth unemployment (.003) (.003) age(young)_trustpa (.006) age(young)_trustinst (.007) age(young)_unemp (.022) age(young)_unra (.003) age(young)_grgdp.012 (.010) Constant (.022) (.027) (.030) (.034) (.034) SD(Constant) (.011) (.012) (.013) (.010) (.014) SD(Residual) (.012) (.011) (.011) (.011) (.011) Observations Groups AIC BIC ICC.03 Robust standard errors in parentheses p < 0.1, p < 0.05, p <

78 Table 4.7 presents results from a multilevel analysis of the 2010-round of the ESS with elite-directed political participation as the dependent variable. The interclass correlation (ICC) value is.03 and this indicates that the country level contributes 3 percent of the variance. This is about 1 percent lower than in the 2008-data. This indicates that the country level explains less in 2010 than in This is in line with the bivariate graphs depicting the relationship between the country level indicators of economic insecurity and elite-directed political participation. Which indicated that there is no relationship between the country level variables and the dependent variable. The null model only includes the dependent variable. Here the constant is positive and significant with a value of.815. This does not say anything as there are no predicators in the model. Model 1 includes age (young), age (old), gender, education and religion into the regression. The significance and direction of the control variables is as expected from theory. Men are more likely to participate in elite-directed political activities than women, all other variables held constant. The strength of the relationship is a bit stronger than in The higher the education of the respondent is, the more likely it is that the respondent will participate in elite-directed activities. The strength of the relationship is a bit weaker than in Religious respondents participate less than non-religious respondents in elite-directed political participation. The strength of the relationship is a bit stronger than in The direction and significance of all three relationships is the same as in When we go from adult (0) to young (1) on age(young) elite-directed political participation decreases, all other variables held constant. This is significant on the one per cent level. This means that the respondents between 18 to 29 participate less than the adults in the 30+ age group. The direction and significance of age (young) is the same as we observed in However, the strength of the relationship has weakened to some extent. This indicates that the age gap in elite-directed political participation between the young people and adults is still present, but to a smaller degree than in The elite-directed political participation increases when moving from (0) to 60+(1) on age(old), all other variables held constant. This means that the respondents that are 60 years and over participate more than the respondents that are under 66

79 60 years old in elite-directed political activities. The direction and significance level is in line with what was observed in the 2008-data. However, the strength of the relationship increases from 2008 till This indicates that the age gap in elite-directed participation between those under 60 and those over 60 is still prescient, and it is growing. What we can infer from this is that there are clear age gaps in elite-directed political participation. Young respondents between years participate significantly less than respondents over 30 years. And the respondents under the age of 60 participate less than respondents over the age of 60 in elite-directed activities. This means that elite-directed participation increases with rising age. The age gap is in line with what we would expect if hypothesis 1 was confirmed. However, I need to compare this finding to the findings on elite-challenging participation in order to test this hypothesis. Model 2 includes the above mentioned variables, individual level unemployment and two measures for political trust. When going from having a job to being unemployed there is a significant decrease in elite-directed political participation, all other variables held constant. The significance level and the direction of the relationship is the same as was observed in the 2008-data. However, the strength of the relationship increases. This means that not having a job decreases the elite-directed political participation. This indicates support for a civic voluntarism view on the individual level as it appears that becoming unemployed decreases elite-directed political participation. It appears that one needs resources in order to participate in elite directed political activities. Which is in line with what Verba et. al. (1995) claim. A units increase in trust in political authorities leads to a increase in elite-directed political participation, all other variables held constant. This means that higher trust in political parties and politicians leads to higher levels of elite-directed political participation. The direction, strength and significance level of the relationship is in line with what was observed in the 2008-data. This lends support to the idea that trust in political authorities has a motivating effect on elite-directed political participation. Trust in institutions is not significant, but has the same direction as trust in political authorities. The direction of trust in institutions is a change from 2008 as it previously had a negative direction. All of 67

80 the independent variables and the control variables from the previous models have the same direction, strength and significance level as before. In model 3 the country level variable are added to the regression. These are: real GDP growth rate and change in youth unemployment rate. The direction of the real GDP growth rate is negative. Which means that a increase in the real GDP growth rate leads to a decrease in political participation. The direction of real GDP growth rate is the opposite of what was observed in the 2008-data. The strength has also increased a bit. However, the coefficient is still not significant. The direction of the change in youth unemployment rate variable is negative which is in line with what we saw in the 2008-data. Which means that a rise in the change in youth unemployment rate leads to a decrease in political participation. The strength of the relationship is also much weaker. However, the change in youth unemployment rate is still insignificant. The individual level independent variables and control variables still have the same values and the same significance level. The results from the two previous models indicate that real GDP growth rate and change youth unemployment does not have an significant effect on elite-directed political participation. This strengthens the findings from the 2008 data and indicates that individual level variables have more of an impact on elite-directed political participation. In model 4 I add different interactions between age(young) and the different independent variables. This is in order to start testing the last two hypothesis. The individual level variables still have the same strength, direction and significance level. The significance level of the real GDP growth rate has changed in this model and it is now significant. The strength and direction is still the same. One units increase in the real GDP growth rate leads to a decrease in elite-directed political participation. The change in youth unemployment rate variable is still not significant. Of the five interaction terms only one of them is significant. This is the interaction between age(young) and trust in political authorities. This interaction term was also significant in the 2008-data. However, the significance level has changed from the.01-level to the.10-level. The interaction term still indicates a negative relationship as it did in The strength of the relationship has also decreased from Below I present 68

81 a margins plot of the interaction term, as suggested by Bambor et al. (2006). Figure 4.4: Interaction effect between age (young) and trust in political authorities Figure 4.4 illustrates the effect of age (young) conditioned by trust in political authorities on elite-directed political participaiton in Going from an adult to a young person there is a fall in political participation and the difference increaces by the level of trust. Young people with a low degree of trust in political participation have a lower degree of political participation than adults with low degrees of trust in political authorities. The age gap is still present, but it appears to be larger than in The figure shows marginal effects controlled for the other variables in the regression. 69

82 Table 4.8: Multilevel regression with elite-challenging participation as dependent variable, 2008 (0) (1) (2) (3) (4) Elite chall. Elite chall. Elite chall. Elite chall. Elite chall. Age(Young) (.021) (.020) (.020) (.052) Age(old) (.019) (.020) (.020) (.020) Gender (.019) (.019) (.019) (.019) Education (.013) (.014) (.014) (.013) Religion (.021) (.020) (.020) (.020) Unemployment (.016) (.016) (.015) Trust in political authorities (.004) (.004) (.004) Trust in institutions (.006) (.006) (.005) Real GDP growth rate (.040) (.040) Change youth unemployment (.015) (.015) age(young)_trustpa.007 (.007) age(young)_trustinst (.010) age(young)_unemp (0.0262) age(young)_unra (.005) age(young)_grgdp (.009) Constant (.053) (.047) (.055) (.081) (.082) SD(Constant) (.023) (.034) (.034) (.039) (.039) SD(Residual) (.032) (.032) (.032) (.032) (.032) Observations Groups AIC BIC ICC.11 Robust standard errors in parentheses p < 0.1, p < 0.05, p <

83 In table 4.8 results from multilevel analysis of the 2008 round of the ESS with elitechallenging political participation as the dependent variable are presented. The interclass correlation (ICC) value is.11 and this indicates that the country level stands for 11 per cent of the variance. This is in line with what we saw in the bivariate graphs between elite-challenging political participation and country level variables connected to economic insecurity. The null model only includes the dependent variable. Here the constant is positive and significant with a value of.424. This does not say anything as there are no predicators in the model. In model 1 age(young), age(old), gender, education and religion is added to the regression. Age (young) is not significant and it has a negative direction. If it was significant this would indicate that going from 30+(0) to (1) would decrease elite-challenging political participation. This is not what was expected and it gives some support to the idea of Millner (2010). The difference between the young people and the group in the middle appears to be negligible. Age (old) is significant and the direction is negative. When going from (0) to 60+(1) the elite-challenging political participation decreases, holding all other variables constant. This indicates that elite-challenging political participation has a curve linear relationship with age. The group between take part in the elite-challenging political participation to a larger extent than the group over 60. So does the middle-aged group between 30 to 59, who take part in elite-challenging political participation to a larger extent than those 60 and older. There is really no significant difference between the young people and the middle group. Although there is a age gap in elite-challenging political participation. It is not between the young and the adults. It is between those below and above 60. This indicates that those below 60 have a larger likelihood of taking part in elite-challenging activities than those above 60. It confirms the findings from the bivariate analysis. All of the control variables have significance levels and the direction that is expected from theory. Gender is significant and has a positive direction. This means that women take part in elite-directed political participation to a larger extent than men. One units rise in education leads to a rise in elite-directed political participation, all other variables held 71

84 constant. The higher the educational level of a respondent has the more he or she participates through elite-challenging political activities. This is in line with the theoretical expectations as rising levels of education gives higher degrees of elite-directed political participation. Elitechallenging political participation increases with having a religious belonging. This is in line with the theoretical expectations. In model 2 trust in political authorities, trust in institutions and unemployment is added in addition to the variables already included in the regression. Unemployment has a significant impact on elite-challenging political participation. When going from employed to unemployed there is a decrease in elite-challenging political participation, all other variables held constant. This might indicate that loosing resource has a negative impact on elitechallenging political participation. This is in line with the civic voluntarism model and goes against the predictions of grievance theory. However, this only looks at the individual level factors and it could be that adding interactions gives another picture. So it is to early to conclude about hypothesis three. There is a positive relationship between trust in political authorities and elite-challenging political participation. The direction is the same as with elite-directed political participation. However, the relationship between the two is insignificant. Trust in institutions and elite-challenging political participation has a negative and significant relationship. One units increase in institutional trust gives a decrease in elite challenging political participation, all other variables held constant. This indicates that a lack of trust in societal institutions could drive and motivate elite-challenging political activities. This is in line with the findings of Hooghe and Marien (2013). This type of relationship is what is expected in hypothesis two. However, the hypothesis is not tested until model 4. In model 3 I look at how country level indicators affect elite-challenging political participation. All individual level indicators have the same direction and have the same significance level as in the previous models. Real GDP growth rate has a negative relationship. One units growth in the real GDP growth rate would lead to a fall in elite-challenging political participation. However, the relationship is not significant. Change in youth unemployment rate has a very weak positive effect on elite-challenging political participation. One units 72

85 growth in the youth unemployment rate would lead to an increase in elite- challenging political participation. However, it is not significant. This model presents some interesting findings when it comes to hypothesis 3. It appears that country level variables of economic insecurity does not have a significant impact on elite-challenging political participation. This might indicate that there is no relationship there. However, before the hypothesis can be rejected I need to look at a model with interactions to evaluate this. In model 4 I include all of the same variables as in the previous model, but here I include interactions for each of the independent variables with age (young). The individual level indicators still have the same direction and are all still significant. The value of age (young) drops to some extent, but this is due to the fact that it is included in all the interactions. The country level indicators still have the same direction, strength and they are not significant. Only two of the interactions are significant. These are the interactions between age(young)*real GDP growth rate and age(young)*change in youth unemployment rate. The interaction between age(young) and change in youth unemployment rate is negative. The interaction between age (young) and real GDP growth rate is also negative. This confounds the picture on the one hand a increase in youth unemployment rate indicates economic hardship and this leads to lower political participation among young people. On the other hand a increase in real GDP growth rate which indicates economic growth leads to a decrease in political participation. An interesting finding is that the interactions between both the interactions between trust and elite-challenging political participation are insignificant. In both table 12 and 13 the interaction between trust in political authorities and elite-challenging political participation are significant. This is in contrast to what I expected beforehand and indicates that there might not be hold for hypothesis 2 in the data. 73

86 Figure 4.5: Interaction effect between age (young) and change in youth unemployment rates and between age (young) and real GDP growth rate In figure 4.5 I present two graphs. 21 The one on the left illustrates the negative effect of age (young) decreases with rising real GDP growth rates. Young people living in a country with a low growth rate in the real GDP tend to participate more in elite-challenging political participation than young people in countries with high growth rates in the real GDP. This supports a grievance perspective on political participation among young people. However, an important thing to point out is that the interaction is not significant for the lowest values of the real GDP rate. This can be observed from the crossing lines in the marginal effects graph. The figure to the right shows that political participation among young people falls with a growing change in the youth unemployment rates. This indicates that young people living in a country with a decrease in the youth unemployment rate participate in elite-challenging political participation to a larger extent than young people living in a country with a increase in the youth unemployment rate. Overall I conclude that this is a null finding. We can t say if the grievance perspective or the civic voluntarism resives support. 21 The figures show marginal effects controlled for the other variables in the regression. 74