How media contexts shape political support: Individual and macro effects

How media contexts shape political support: Individual and macro effects Bouchra Arbaoui Wouter van der Brug Philip van Praag University of Amsterdam, Department of Political Science ASSR and ASCoR April 2009 First draft Do not cite without permission Contact address Bouchra Arbaoui University of Amsterdam, ASSR Kloveniersburgwal 48 1012 CX Amsterdam the Netherlands e-mail: B.Arbaoui@uva.nl 1

Introduction The nature and extent of government regulation of the media varies substantially across countries (Bennet 1998, Hallin and Mancini 2004). Yet, deregulation of the media market is a common trend among European countries. Technological change is believed to be the main factor behind this trend, as this conceivably puts pressure on policy makers to deregulate media markets. As a result, the media systems of most European states have become, to varying degrees, more commercialised. In the television sector, the consequences of these trends are most visible, as the previous state monopoly on television has been lifted throughout Western Europe. The television sector in Europe has been commercializing and public television is increasingly exposed to competition with commercial TV. As a result, it risks losing its original political and educational objectives. This paper analyses the consequences of commercialisation and increased competition of the television sector for political attitudes and, in particular, political support. In the relevant literature, two opposing theories prevail on the effects of media (exposure) on political support. Some scholars argue that the media can mobilize and inform citizens (e.g., Norris, 2000; Dalton, 2002) and others argue that, on the contrary, the media contribute to political cynicism and alienation (e.g., Putnam, 2000). While these views are contradictory, both of them might apply according to the specific media context. In most comparative studies on media effects at the micro (individual) level, TV exposure is taken as the only explanatory variable. The nature of macro-environmental flows of political information and the cross-country differences in media systems is frequently ignored. While the media context might not influence the patterns of media exposure (it is rather the other way around), it might influence the quality of the information that people are exposed to. 2

While acknowledging the agency of audiences, the variation in media effects across different social contexts and countries cannot uniquely be explained by individual characteristics. Individual characteristics only serve as a filter for structural communication patterns determined by the specific media systems. Media use is not only affected by individual characteristics but also by structural factors defining the media environment in which the individual and their social networks live. The individual is embedded in the structural context of media supply and outlet shaped by the specific media system. In addition to media exposure, we therefore need to understand how media systems are constructed. This paper argues that contextual media variables should be included in addition to individual variables in order to attain a more comprehensive understanding of media effects on political support. To this avail, we develop media system variables for 17 European countries, which include measures for media-party parallelism, the degree of commercialism and competitiveness within the television sector. In this way, we focus on the media market characteristics instead of characteristics of a single communication medium. Drawing on data from the European Social Survey 2004-2005 and employing multilevel regression analysis, we test whether these contextual variables affect political support, and whether and how the effects of individual level variables are moderated by these contextual variables. This paper contributes to the existing literature in two ways: First, it shows the theoretical relevance of contextual media system variables for assessing media and media exposure effects on political attitudes. By doing this we theoretically clarify the media malaise thesis and go beyond the effects of media exposure only. Second, this study employs systematically collected measures of commercialism and degree of competition in the television sector, as well as a measure of media-party parallelism for 16 European countries allowing for cross-sectional comparison. By focussing on the (cross-) interaction of individual factors and contextual media factors we connect the theoretical arguments that media (exposure) has a considerable effect on political attitudes with 3

empirical studies that often show minimal media (exposure) effect and even contradictory results. We start by defining our dependent variable, that is, political support, which is based on Easton s work. Subsequently, we explain the processes through which the media affect political support and identify and define the structural and individual factors that are hypothesized to determine these processes. By doing so, we aim to explain the causal mechanisms by which commercialisation and increasing competition of the media sector can theoretically be expected to affect political attitudes and, particularly, political support. In addition, we explain how these structural factors interact with individual characteristics and what effects these interactions have for individual s political support. In the third section we empirically test our theoretical assertions by using a multilevel model and present our empirical results and conclusions. 1. Political support and media influence Political support entails more than an attitudinal component, but also a behavioural one. Support, refers to the way in which a person evaluatively orients himself to some object through either his attitudes or his behaviour (Easton 1975: 436, emphasis added). From this definition we can distinguish two forms of support. The first one is behavioural and the second one is attitudinal. Easton (1965) calls the first one overt support and consists of supportive actions; covert support is the second form and refers to sentiments towards political objects. Covert support is further distinguished in specific and diffuse support. As this paper is focused on individual feelings towards the political system, we will limit our further analysis to specific and diffuse support. 4

While specific support involves favourable attitudes in response to outputs that satisfy members policy demands, diffuse support involves attitudes of general good will, independent of immediate outputs (Easton 1965 in Muller 1970:1151). Diffuse support refers to citizens attitudes towards the political systems as a whole and not only to temporary political situations and outcomes. Specific support is contingent on the authorities presumed behaviour; it is object-specific and refers to satisfaction with political outputs and performances. According to Easton, diffuse support consists of sentiments of trust and legitimacy (Easton 1975: 453). While trust refers to the attitudes towards the political object, the core of legitimacy constitutes of values. Legitimacy does not depend directly on the performance of political authorities. Empirically, it is difficult to make a clear distinction between specific and diffuse support because confidence/trust also contains an element of evaluation. This evaluation might not only be rational but also influenced by affective feelings. In addition, people often confuse the evaluation of a system or institution with the evaluation of its current incumbents and its current outcomes. However, various studies analysing cross-national longitudinal studies through factor analysis (Dalton 2004, Klingemann 1999) have shown that distinct dimensions of support exist empirically, even though they often overlap considerably. Muller and Jukam (1977) have also demonstrated the possibility to distinguish incumbent evaluation from system evaluation, since the measures for incumbent evaluation correlate differently with ideology than measures evaluating systems. Muller (1970) argued that people do not need to match their specific policy demands to outputs to be able to evaluate the authorities but they might use a general evaluation of a particular issue that they find important like domestic or foreign affairs. In line with Muller (1970), we will assume that specific support is best understood by looking at citizens general evaluation of authorities and not at specific demand satisfaction. 5

In sum, political support is generated by socialisation processes, direct experience with politics affecting the individual perception of political outcomes and the degree into which the individual political demands are presented and taken into account. Easton (1965) argues that in political system authorities and citizens engage in a continuous learning process and exchange information about political demands and political outcomes. In modern societies, the mass media play an important role in this learning process. Very few people have direct contact with politicians, particularly national politicians. So, people learn about politics through the media. They identify through media what social-political problems are and what the political answers are for these problems. They learn to evaluate if their political demands are taken seriously and learn via media and social networks what political outcomes are and how to evaluate them. On the other hand, political authorities learn also from media how citizens evaluate political outcomes and therefore anticipate the amount of specific support. This in its turn generates new political demands and eventually political outcomes. During this learning process an information distortion occurs (Easton 1965). This dysfunction affects the generation of support because it influences the perception of outputs and also political authorities anticipation on the system members wants and (specific) support. This information distortion or bias is inevitable as communication processes are by nature complex, and since power is not distributed equally within society, the idea of the media functioning as a mirror of society remains purely imaginary. A bias seems thus inevitable in mediating processes. This bias might however be in favour of the political system if it is a partisan political bias or against it if it is negative. These processes occurring during the communication process between political authorities and citizens described above are known as process of agenda setting and framing 1. How these processes affect political support is shown in Figure 1. 1 See for instance Scheufele (2000) describing these processes and the interrelationship between framing and agenda setting. 6

Figure 1. Agenda setting and framing impact on political support 2 : Now that we have defined the processes through which media is likely to affect political support, the following question that needs to be answered is: Which factors at the media system level determine the way politics are framed and the extent to which agenda setting by media occurs. In addition, we wish to determine the individual and structural conditions that structure processes of framing and agenda setting. In the next section we try to identify the relevant contextual factors as well as their interactions with individual characteristics. 2. Contextual and individual factors conditioning the nature and effect of framing and agenda setting processes Processes of professionalisation of journalism in Europe went hand in hand with the process of marketising the media field. A decline of political instrumentalism has made room for a market mentality in which perceived public demands are the guiding line. According to some scholars, news selection is increasingly guided by the slogan good news in no news leading to a tendency to covering negative news more then positive news. In addition, commercialism has led to homogenisation so that media provide 2 It would be more adequate to speak of perceived demands representativity. 7

increasingly similar political coverage (e.g. Jamieson & Campbell 1998; Picard 1998). Media political coverage is characterized by tabloidisation, entertainment and superficiality in order to attract as much audience as possible. For instance, researchers have identified various frames in the US press: Strategic game framing (Cappella and Jamieson 1997); Horse race framing (Benoit et al. 2005); and Episodic framing (Iyengar 1991); Conflict framing (Brants and Kempen 2002). These frames have an impact on political orientations of citizens as they depict politics as a game and do not concentrate on issues. Patterson (2000) argues that commercialization of the media is the explanatory force of the growing tendency of interpretative journalistic style and framing politics as a game. This seems to encourage political cynicism and negativity and, hence, affects political support. Empirical studies (e.g. Esser et al, 2001) of strategic framing and meta-coverage show that the degree of pervasiveness of this type of reporting varies. Election periods seem to generate more of such coverage than non-election periods (Esser et al. 2001), while liberal media systems seem to produce higher levels of such reporting than corporatist models (De Vreese 2008: 7). In their cross-sectional comparative study on the formation of campaign agenda setting in the UK and the US, Semetko et al (1991: 178) argue that more commercial systems are associated with more power for journalists to set the political agenda and that more competition is associated with more attention to perceived audience interests and less attention to politicians agenda 3. Framing politics in terms of strategy or conflict is not only a consequence of commercialisation. The extent to which this happens can also be expected to depend upon the nature of political systems, political culture and, hence, media systems. Various factors influence how journalists frame politics: social norms and values, organizational pressure and constraints, pressures of interest groups, journalistic routines, as well as ideological and political orientations of journalists (Scheufele 2000: 307). The effects of 3 Blaming market mechanisms and commercialism and competition in particular is not new. The debate goes back to early 19 th century. Much of the studies echo this old debate and Habermas who believes that the political debate becomes less rational, sensational and superficial as a consequence of commercialisation, arguing that the press becomes more manipulated when it becomes more commercialised (1989: 185). 8

framing might be moderated by the extent to which media landscapes are diverse and plural. In a highly pluralist media context, it is likely that the effect of one media message is counterbalanced by another, thus resulting in almost neutral media effects (e.g.zaller 1992). Competition is believed to be good for stimulating pluralism in the media landscape and might therefore moderate the effects of commercialism. Increased competition associated with technological development might however stimulate the dominance of certain frames which have proven to attract the largest audiences. When competition becomes hyper-competition it may create media that seek to survive producing bad quality media coverage and focussing on entertainment in order to guaranty a minimum level of audience. In such cases, competition might strengthen the negative effects of commercialism. In addition, internal and external pluralism are the two dimensions through which media systems handle the diversity of political loyalties and orientations (Hallin and Mancini 2004). External pluralism is achieved at the level of the whole media system when media outlets are affiliated with different (party) political groups. As long as each of the groups has their own media, external pluralism is guaranteed, even if each of the outlets is very biased. This would be the case when there is a high degree of party-press parallelism. Internal pluralism is associated with pluralism at the level of individual media outlets. In this case, the medium does not have a particular political leaning but will attempt to present different perspectives on a topic. While broadcast media may be characterised by internal pluralism, the press might be dominated by external pluralism (Hallin and Mancini 2004). Media-party parallelism (MPP) refers to the strength of connections between the media and political actors and the balance between advocacy and neutral/informational traditions of political journalism. MPP is strongest when each medium (press and TV) supports a party, which is visible in the leader columns and in the editorial parts of newspapers but sometimes also in the news items themselves (see for instance Van Kempen 2008). 9

While political parallelism reduces internal pluralism it increases the level of external pluralism. It may also have other advantages as political news might dominate entertainment news if the level of political parallelism is high, which means less superficiality in news. From this, we can hypothesise that a higher degree of external and internal pluralism implies that various frames coexist in the media landscape, indicating that there is no dominant frame. In addition a highly partisan media landscape is likely to reinforce existing citizens partisan orientations, which, in turn, are known to be related to turnout (e.g., Brynin & Newton 2003; Franklin 2004; Gimpel, Dyck, & Shaw 2004). In a study of cross-section survey data from fifteen member states of the European Union in 1999, Van Kempen (2008) shows that (external) media party parallelism has a positive effect on electoral participation, although this effect is moderated by political interest. The effect is stronger for the least politically interested because they are more open to mobilizing influences. The smaller the repertoire and the more limited the information and arguments people have, the greater their readiness to accept new information 4. This effect is stronger for people who are neither involved nor interested in politics (e.g. Cappella, Price, & Nir 2002). This seems to contradict Norris (2000) view. Other studies on media effects also acknowledge that media effects are moderated by individual characteristics, which either restrict or reinforce them (e.g. McQuail 1987; Miller 1991). According to Zaller (1996), peoples attitudes determine their selective use of media and the way they read media messages, so that existing attitudes are only reinforced but not changed by media use. This is indeed a major methodological problem for people studying media effects, since as cross-sectional studies have difficulties in coping with selection bias. However, this view seems to neglect that media outlets are determined by media system characteristics influencing the choices that individuals have in media use. Zaller (1996) seems to assume a high degree of external and internal pluralism in the media landscape. 4 The theory of cognitive dissonance (Festinger, 1957 quoted in van kempen (2008: 21)) holds that people will try to reduce whatever dissonance they experience between attitudes and information. 10

In this view, political attitudes are seen as mainly formed through a socialization process and are correlated with socio-ethnic background, religion, gender and age and personal experience with the political system. However, this view seems to overlook that almost al personal experiences with politics take place through the media and that the media may play an important role in the socialisation process and in forming attitudes too 5. Social networks are believed to determine the effect of media messages. When media messages are not in concordance with the opinion of the social network of the message receivers, the message is not likely to have an effect (see for instance Newton 2006). Yet it is likely that social networks are also subject to the same media context as individuals, so that we still can speak about structural media effects as they also affect the dominant opinions within social networks. The effect of media messages is generally believed to be dependent on political knowledge and the interest that audiences have in the issue and on the importance of the message (e.g. Newton 2006). The more knowledge and cognitive involvement about the issue and the more important the issue in the every day life of a person, the lower the effect the message has. This is because individuals capabilities of detecting bias in the media message will increase. A critical reading of media messages is more likely when political knowledge and interests increases. Also people without political interest and knowledge seem to be relatively immune against media effects because they pay little attention to political coverage (e.g. Norris 2000). This would mean that persons with an intermediate level of political knowledge and interest are most likely to be affected by media political coverage (Zaller, 2002). 5 Almond and Powell (1966) see media as an important player in the political culture as it creates attitudes and orientations and as it affects existing knowledge. Attitudes and orientations Political culture is a set of propensities or orientations towards action in a given population. 11

On the basis of this discussion, we may now define the following hypotheses, which will be tested in this study. H 1: Commercialism has a negative effect on political support. H 2: The negative effect of commercialism is moderated by competition. Low and hyper levels of competition reinforce the negative effects of commercialism and high levels of competition counterbalance them 6. H 3: Negative effects of commercialism are most visible when Media-Party Parallelism (MPP) is low and are counterbalanced by higher levels of MPP. H 4 TV exposure effects on political support are expected to be negative in highly commercial media contexts. H 5: the effect of commercialism is dependent on the individual level of political interests and knowledge: we expect the effects of commercialism to be the highest among people with an average political interest and knowledge. 2. Data and Measures We wish to estimate the effects of media context on political support. This requires individual level measures of political support that can be compared across a large number of media contexts. For the purpose of this study we use the second wave of the European Social Survey, collected between 2004 and 2006. We selected the following 17 countries, for which we were able to collect the relevant media context variables: Austria, Denmark, Finland, Flanders, France, Germany, Greece, Ireland, Italy, The Netherlands, Norway, Portugal, Spain, Sweden, Switzerland, United Kingdom and Wallonia 7. 6 This suggests a non-linear interaction between commercialism and commercialism. 7 Belgium is divided in two regions: Flanders and Wallonia. This is because these two regions have different media systems. 12

2.1. Measures of political support A measure of satisfaction with democracy is theoretically a measure of support for regime principles, norms, procedures and institutions. Because this variable contains an evaluation at the system level and not at the authority level it should be seen as a measure of diffuse support. The wording of this question in the European Social Survey (ESS) is: On the whole, how satisfied are you with the way democracy works in your country? (0 stands for extremely dissatisfied and 10 for extremely satisfied). However, the wording of the question in the ESS is problematic as the question focuses on how democracy works. It is not clear if the respondents focuses on the outcomes of the current democratic government or expresses their feelings toward the democratic system in general. While this measure focuses on the system level it emphasises on performances, therefore it should be seen as measure of both diffuse and specific support. According to Easton, a typical measure of support for political authorities would be the evaluation of the functioning of parliament, political parties and politicians. In the ESS we find questions about trust in these various political authorities. As trust includes affective feelings and rational evaluation, we consider this question as a mixture of specific and diffuse support. The wording of these measures is: Please tell me on a score of 0-10 how much you personally trust (politicians/political parties/country s parliament). 0 means you do not trust an institution at all, and 10 means you have complete trust. For the sake of parsimony we focus in this paper on political trust in Politicians. 13

2.2. Commercialism, competition in the TV sector and media party parallelism McQuail (1986) distinguishes between TV systems (1) entirely financed by private capital in the form of advertising or sponsorship, which are run as businesses without public regulation of content; (2) financed by private capital but functioning in a broader regulated system; (3) financed by direct pay from subscription with or without advertising capital; (4) based on public service broadcasting and financed by public capital and advertising subsidy (the most common one) and (5) where the public TV is entirely funded by public money. McQuail s categorisation is on the basis of the extent to which TV stations depend on commercial funding. Although no private money is involved, he calls the last model a weak version of commercialism which refers to the dilution of cultural and informational content in the interests of maximising audiences (1986:156). The core of al these models is the financing source and the fact that even when a TV company or channel does not rely on private income it might be involved in the market game as they might follow the market rules in order to keep their audience share. We largely follow McQuail s conceptual framework. However, rather than classifying media systems in five types, we measure the degree of commercialism in the TV sector by one single indicator: the extent to which the earnings of TV companies depend on private money. This is computed in two steps. First, we computed for each of the national TV companies its revenues from advertising, sponsoring, and subscription as a proportion of its total earnings. In a second step, we determined the weight of a TV company in the TV market by its audience share. The share of audiences that a TV company has is often neglected in determining if a national TV sector is commercialised. In many cases, only attention is paid to the source of income or the percentage of advertising time. The degree of commercialism in the TV sector is measured by weighting the percentage of commercial income that a TV company gets by its daily share of the audience market. We drew on the yearbook (2007) of the European Audiovisual Observatory for 14

measuring the income of TV companies 8. Only public TV companies receive noncommercial income. Commercial incomes are mainly generated by advertising, sponsoring and merchandising. To calculate our measure, we used the average commercial income and audience share over the years 2004-2005 and 2006. We did so because the individual data is gathered over 2004-2006. Commercialism = F (proportion commercial income, audience share) = 1-Σ noncommercial income TV company * its audience share). Commr = CIC * m = 1 CINC * i i i i i m i Where: CIC i is the percentage of commercial income in a TV i company and m i is its market share. CINC i stands for the non-commercial part of income in a TV company i. We restrict our measure of competition to audiences. The level of competition depends on the number of TV companies and also on their relative market share. If the number of companies is equal in two TV sectors, the market where one channel monopolises a larger part of the audience is less competitive then the market where the audience is more evenly distributed (see also van der Wurff 2004). The degree of competition in a TV sector is measured by summing up the squares of the market daily share for each TV company or organisation and subtracting it from 1 9. We use a Herfindahl-Hirschman Index (HHI) and we recoded the index so that higher values indicate more intense competition (see for instance van der Wurff et al: 2001, 2004). The measure runs from almost 0 to 1, where 1 stands for high competition. Compt = 1 i m 2 i 8 For Greece public TV income we used only the income for 2004. 9 We draw on the MAVISE database of TV companies and TV channels in the European Union of the European Audiovisual Observatory: http://mavise.obs.coe.int. The audience shares for the TV companies can be found in the European Audiovisual observatory yearbooks (2007). We calculated the average of 2004, 2005 and 2006. 15

Where m i stands for the market share of TV Company (i) or organisation (i), the audience market shares are average daily share over 2004-2006. For measuring media party parallelism we rely on van Kempen s (2008) measurement. Media party parallelism is calculated by regressing party preference scores on newspaper- and television news exposure. MPP scores range within [0; 100] where 0 stands for no media-party parallelism and 100 for a maximum media-party parallelism 10. We use individual data from the European Election Study 1999. Figure 2: Countries levels of commercialism, competition and MPP (scale from 0 to 1) 11. 10 For the exact procedure we refer to Van Kempen (2008). 11 Switzerland is missing on the graph because we do not have its MPP scores. 16

3. Models and Empirical results If the aforementioned theoretical assertions have some empirical ground we expect to find that (1) the average political support significantly varies between countries and that (2) TV exposure effects on political support are significantly different from country to country and that (3) the between country variation in the average political support can be partially explained 12 by the above mentioned media context variables: commercialism, competition and media-party parallelism; (4) the effects of individual and contextual media variables are moderated by political knowledge and interests. 3.1. Between-country variation in political support We first we test whether the average political support varies significantly between countries. In other words: Is the country level a meaningful level of analysis? To answer this question, we compare the usual empty OLS regression model and an (empty) random intercept (U 0j ) model with political support as a dependent variable. We restrict the analysis for the sake of parsimony to political trust in politicians (PTP) which is a measure of both diffuse and specific support for political system authorities and satisfaction about how democracy works (SATDEM) which is a measure for political support at the system level. m 00 : PTP = α 0 m : SATDEM 00 + = R α + 0 R m : PTP = α + U + R 0 0 0 j m : SATDEM = α + U 0 0 0 j + R The results show that the average political support varies significantly from country to country (see Table 1). The interclass correlation is 0.12 for political trust in politicians and 0.14 for satisfaction with democracy. Since both interclass correlations are 12 Explanation has a limited meaning here as it means that the values of the dependent variable are predicted by values of the explanatory variables. 17

significant, it is sensible to estimate our effects by means of multilevel modelling. There is a between country variance of 0.63 around the average for political trust in politicians and 0.74 for satisfaction with democracy. Table 1. Between country variation in average Political Support. Estimates PTP SATDEM Average 3.9 5.67 (0.19) (0.2) Between country intercept variance 0.63 0.74 Within country individual variance 4.6 4.67 Deviance (empty)random intercept model (m 0 ) 136133.49 134068.21 Deviance empty OLS model (m 00 ) 140028.34 138518.3 Interclass correlation 0.12 0.14. 3.2. Media exposure effects on political support After having empirically shown that there are significant differences between countries in the level of political support, we test whether media exposure effects on political support are significantly different between countries. Media exposure is measured by the amount of total hours television watched (TVTOT) and the amount of hours political news are watched (TVPOL) (for the exact wordings of these measures see appendix). For this purpose we estimate the random intercept random slope (U ; i#1) model (m 2 ): Model(m SATDEM ) 2 PTP : = α + β TVTOT + β TVPOL 0 1 2 + U 1 j TVTOT + U 2 j TVPOL + + U R = α + β TVTOT + β TVPOL 0 1 2 + U 0 j + U 1 j TVTOT + U 2 j TVPOL 0 j + R 18

We compare this model with the random intercept model only which assumes that the effects of media exposure are not significantly different between countries. Model ( m ) : 01 PTP SATDEM = α 0 + β TVTOT 1 = α 0 + β 1 TVTOT + β TVPOL + U + R 2 0 j + β 2 TVPOL + U 0 j + R We first compare the random intercept only model with the random intercept random slope for TV watching news only. This random slope was significant, which means that the amount of TV news watching has a significantly different effect among countries. We subsequently compare the random intercept model with random slopes for both media exposure variables with this model. We find that both random slopes of both media exposure variables are significant. This means that the length of TV exposure affects political support in a varying extent depending on the country. For the results of both models see Table 2. From this, we can conclude that the effects of the amount of news watching on individual political support are generally positive. We can also conclude that the effects are substantially (and significantly) different in the countries that we study. 13 13 See for instance Snders and Bosker 1999 for interpreting the slope variance. 19

Table 2 14. Fixed and random effects of TV exposure variables on political support Political trust in Politicians (PTP) Satisfaction with how democracy works (SATDEM) Intercept 3.92 (0.19) 5.67 (0.2) Fixed effect TVPOL 0.1 (0.01) 0.07 (0.01) Fixed effect TVTOT -0.06 (0.017) -0.06 (0.02) Between country intercept variance 0.61 0.71 Random slope variance TVPOL 0.0007 0.0006 Random slope Variance TVTOT 0.004 0.004 Covariance intercept/slope TVPOL -0.007 0.008 Covariance intercept/slope TVTOT -0.02 4.6 Within country individual level variance 4.57 4.62 Deviance random intercept random slopes both media exposure variables model (m 2 ) Deviance random intercept random slope TVPOL variable model (fixed effect TVTOT) (m 1 ) (df (m 1-131940.319 (χ2 (m 2 -m 1 ) =54.43; df (3) P=0.000) 131994.749(χ2 (m 1 -m 01 ) =20.01; df (3) P=0.0001) 129906.318 (χ2 (m 2 -m 1 ) =57.017; df (3) P=0.000) 129963.335 (χ2 (m 2 -m 1 ) =10.62; df (3) P=0.0139) Deviance random intercept model, fixed media exposure effects (m 01 ) 132014.76 129973.962 14 Al variables are centered random their individuals mean. The intercept is thus political support for an individual with an average TV exposure. The results of the model with fixed total TV watching (m 1 ) are not reported. In the table we report only the model deviance. 20

The main effect of total TV watching duration in general on both measures of individual political support is positive. However, the variance across countries is so large, that the effects may even be negative in some countries. The negative covariance between both slopes and intercepts means that countries with higher level of political support (for an individual with an average TV exposure) have a lower within-country effect of length media exposure. In other words, the length of media exposure tends to have more (within) effect in countries with a lower average level of political support (for an individual with an average TV exposure). The presented empirical evidence in this section suggests that the more people watch TV, the more likely it is that they have lower support for their political system. However, the more they watch news, the more likely people are to have higher support for their political system, although this effect varies across countries. The question that still needs to be answered is: can the degree of commercialism, competition and media party parallelism explain these differences in the effect of length TV exposure on political trust? One of the explanation of the negative effects of total TV watching as predicted by the video malaise hypothesis is that the negativism and superficiality of politics in TV (especially News coverage) are responsible for this. As we have shown, the (within and between) effect of news watching are positive but the effect of length total TV watching is in some countries negative. Does this mean that TV news coverage is not negative but other TV programmes are? If other TV programmes are negative and politically superficial does that mean that the more a person watches these programmes the more they are negatively affected by it? We tested whether the degree of commercialism, competition and media partyparallelism explain some of the between-country variation in (length of) TV exposure effects. The results (not reported here) show that, while controlling for commercialism and competition/media-party parallelism, the between-country variation in the effects of TV exposure variables persists. We believe that if TV content tends to be negative generally, it does not matter how long people watch TV but whether they watch TV, as 21

half an hour is enough to poison you. In addition media might even affects people who doe not watch TV. Various studies argue that media has effects also through social networks and interpersonal communication (e.g. Lenart: 1994). In short, looking at the length of TV watching in general might not reveal media effects as the evidence we find might also mean that people that are more supportive for the political system tend to watch more news and that people who are not supportive for the political system watch more TV in general 15. The between country variation in the effects of TV exposure on political support might also suggest that some other country or individual variables are at work here. The theory on media effects emphasises the moderating effect of political knowledge and interest. Before testing the moderating effect of this variable we first investigate if media context variables explain the between country variance in political trust and thus some of the total individual variance of political support. 3.3. Media system variables effect on political support This section explores if commercialism, competition and media-party parallelism explain part of the variance between countries and thus part of the total individual variance in the level of political support. We test for this purpose the following model: Model(m ) : 02 PTP = α + γ COMR + γ COMP + γ MPP 0 01 02 03 + γ COMR * COMP + γ COMR * MPP 04 05 + U 0 j + R SATDEM = α 0 + γ COMR + γ COMP + γ MPP 01 02 03 + γ COMR * COMP + γ COMR * MPP 04 05 + U 0 j + R 15 Norris 2000 suggested the same interpretation. 22

Because the number of countries in our sample is only 17 we test this model in two steps. First we take the same model but without including media-party parallelism (model: m 021 ) and second we include media party parallelism an exclude competition (model: m 022 ).We compare both and conclude which variable seems to more moderate the effect of commercialism. The results are shown in Table 3. When we tested model m 021 for political trust in politicians, the explanatory power of commercialism and competition was significant (see table 3). The unexplained variance between countries in the level of political trust in politicians decreased with more then 30 percent. The effect of commercialism and competition on individual political trust in politicians is significantly different from zero 16. The interaction between commercialism and competition is like predicted positive but does not add to the model fit and is also is not significant. 16 Significance level is at 0.05 if not otherwise mentioned. 23

Table 3. The effect of commercialism and the moderating effect of competition /MPP 17 Estimates PTP m 021 PTP m 022 SATDEM Average 3.89 (0.17) 3.9 (0.22) m 021 5.55 (0.19) SATDEM m 022 5.56 (0.23) Effect commercialism -4.03 (1.77) -2.94 (2.06) -3.2 (1.8) -3.22 (2.1) Effect competition 4.21 (1.77) - 3.41 (2.08) - Effect MPP - 0.05 (0.06) - 0.02 (0.07) Effect COMR*COMP 7.41 (17.36) - 32.75 (20.4) - Effect COMR*MPP - -0.09 (0.42) - 0.53 (0.44) Between country variance 0.41 0.55 0.56 0.6 Within country individual variance 4.6 4.69 4.67 4.72 Deviance 136126.269 P=0.07 127496.86 P=0.00 134063.8 P=0.22 125300.33 P=0.00 Deviance model without interaction 136126.45 P=0.03 127496.9 P=0.00 134066.18 P=0.36 125301.73 P=0.00 Interclass correlation 0.08 0.1 0.1 0.1 Deviance (empty)random intercept 136133.49 134068.21 model Interclass correlation (m 0 ) 0.12 0.14 *Table entries are ML estimates (the mixed procedure SPSS 15.0). This suggests that the interaction effect between commercialism and competition contrary to what was expected is not empirically relevant. We tested the same model without 17 The number of countries in models with MPP as an explanatory variable is only 16 as the level of mediaparty parallelism in Switzerland is missing. The probabilities of differences in the deviances test are based on the comparison of the model with the empty model. We use a one sided test at the 0.05 level if not otherwise mentioned. 24

interaction and we found as predicted by theory, the effect of commercialism is negative and is equal to -3.9 (1.59). The effect of competition is positive and equal to 3.9(1.63). These effects are significant. We conclude that the degree of commercialism in the TV sector has a negative effect on political trust in politicians but that the degree of competition in contrary has positive effects on Political trust in politicians. When testing model m 021 for satisfaction with how democracy works, we found that although the effect of commercialism is in the expected direction, the model does not perform significantly better then the empty model. We also tested models with the degree of commercialism or the degree of competition in the TV sector and reach the same conclusion. This suggest that if there are negative effects of commercialism in the TV sector on political support, they are limited to support for authorities and do not reach the system level 18. While controlling for MPP, the effect of commercialism on political trust in politicians is still negative and significant at the level 0.1. The effect of MPP is as expected positive although not significant. The interaction between MPP and commercialism is contrary to our expectation negative which means that MPP might strengthen the negative effects of commercialism. But this interaction is not statistically significant. When we looked at the explanatory power of commercialism while controlling for media party parallelism (m 022 ) the fit of the model explaining satisfaction with democracy was significantly much better but the interaction between commercialism and MPP does not significantly contribute to the model fit. We therefore estimated the model again without this interaction first and then because the degree of commercialism did not seem to have effects on satisfaction with democracy we run the model with only MPP as explanatory variable and we discover that this model was as good as in terms of data fit as the model with commercialism as an extra explanatory variable. We conclude that when controlling for media party parallelism, commercialism effects on satisfaction with democracy does not seem to matter. 18 As the literature suggests the non-linear interaction between commercialism and competition we did test this and we reach the same results. When we test the non- linear interaction effect of commercialism and competition on political support the model fit worsen. Therefore we consider this interaction linear. 25

Because the interaction effects seem empirically not to be different from zero, we tested the effects of commercialism, competition and MPP simultaneously and reach the results in Table 4. Table 4. Commercialism, competition and MPP effect on Political support Estimates PTP SATDEM Average 3.94 5.7 (0.16) (0.2) Effect commercialism -4.46 (1.93) Effect competition 3.77 (1.73) Effect MPP 0.035 (0.047) -3.74 (2.35) 1.75 (2.1) 0.07 (0.05) Between country intercept variance 0.42 0.63 Within country individual variance 4.69 4.7 Deviance 127492.751 125301.053 Interclass correlation 0.08 0.11 Deviance model without commercialism and competition 127498.866 125303.483 Deviance (empty)random intercept model 136133.49 134068.21 Interclass correlation (empty)random intercept model 0.12 0.14 The results show that commercialism has a significant negative effect on both indicators of political support, while competition and MPP both have a positive effect. Two of these six effects do not reach statistical significance: the effect of MPP on trust in politicians, and the effect of commercialism on satisfaction with democracy. The other effects are statistically significant at p <.05 (one sided). 26

These contextual variables contribute more to the explanation of trust in politicians than on satisfaction with democracy. The three variables explain about 31 percent of the between country variability of political trust in politicians, and only 15 percent in the country differences in satisfaction with democracy. 19 3.3. Media system variables, TV exposure duration and the moderating effect of political interests The literature on media effects stresses the role of political interests and knowledge as important moderator. The question that we want to answer is whether media context effects on political trust in politicians still persist when we control for political interests and can we find some relevant cross-level interaction between these variables and individual level of political interests, indicating that the level of political interests is indeed a relevant moderator for media effects. To investigate this we tested the following model: PTP = α 0 + β 1 TVTOT + β 2 TVPOL + β 3 POLINT + β 4 POLINT * TVTOT γ COMR + γ COMP + γ MPP + + β COMR * POLINT + β COMP 01 + U 1 j TVTOT 02 02 6 7 + U 2 j TVPOL + U 0 j j + R + β5 POLINT * TVPOL * POLINT + β MPP * POLINT 8 The effect of political interests on political trust in politicians is negative and significant. This means that the more the individual is politically interested, the less political trust in politicians he/she has. The effects of commercialism and competition do not seem to be moderated by the individual level of political interests as there cross-level interaction is not significant. Competition instead is moderated by the individual level of political interests. The effect of competition on political trust for an individual with a minimum political interest is 2.14, for an individual with an average political interest this effect is 2.7 and for an individual with a maximum political interests this effect is 3.37. 19 The model with commercialism and competition as explanatory variables is significantly better than the model with only MPP (P=0.047). The effects of Commercialism and competition are significant a p <.05 one tailed. 27

Table 5. The moderating effect of political interests Estimates PTP Average 3.9 (0.16) Effect commercialism -3.18 (1.75) Effect competition 2.7 (1.5) Effect MPP 0.077 (0.04) Effect POLINTR -0.51 (0.015) TVPOL -0.028 (0.016) TVTOT -0.0055 (0.016) Commercialism*POLINTR -0.24 (0.18) Competition *POLINTR 0.35 (0.15) MPP*POLINTR 0.0046 (0.0045) TVPOl*POLINTR 0.01 (0.01) TVTOT*POLINTR -0.004 (0.008) Between country intercept variance 0.42 Within country individual variance 4.45 Random slope TVPOl Variance 0.0016 Random slope TVTOT Variance 0.003 Covariance random slope intercept TVPOL 0.0016 Covariance random slope intercept TVTOT 0.0007 Deviance 122592.63 Interclass correlation 0.086 Deviance model with media system variables only 127492.75 Deviance (empty) random intercept model 136133.49 Interclass correlation (m 0 ) 0.12 28

The variance of both slopes of TV exposure variables is still significant and the interaction between political interests and TV exposure variables is not significant. This means that the individual level of political interests does not explain the between country differences in the effects of (the length) TV exposure on political trust in politicians. While there was in this model only one cross-level interaction with political interests that was significant, we tested if the effect of individual political interests varies between countries. This is equivalent with testing if political interests have a random slope. We found (results not shown) that this is indeed the case. This means that the effect of political interest varies among countries and that this variation is not explained by the country level of commercialism, competition and media party parallelism, which in turn indicates that we need further analyses that control for other individual and county level variables. In conclusion, the negative effects of commercialism in the TV sector on political trust are not moderated by the individual level of political interests. The positive effects of competition are strengthened by the individual level of political interests. The effect of political interests on political trust in politicians varies between countries but this variation is not explained by our media context variables. The between country variation in the effect of TV exposure duration on political trust in neither explained by individual level of political interests neither by the three media context variables. 29