Introduction

The social impact of stereotyping in television content has been a topic for research for quite a while. It has been found that ethnic minorities and immigrants are routinely associated with criminal (Dixon and Linz, 2000; Iyengar and Gilliam, 2000), economic (Eschholz, Bufkin, and Long, 2002; Jacobs, Claes, and Hooghe, 2015) and cultural threats (Guterman, 2013; Wirtz, Van der Pligt, and Doosje, 2016). The expectation therefore is that television consumption and frequent exposure to this type of messages will be positively associated with negative outgroup attitudes in society (Entman and Gross, 2008; Schemer, 2014). In this line of research, however, it is important to distinguish between various forms of media content. Previous research has demonstrated that the relationship between attitudes and, for example, entertainment television can be totally different from seeking information on television (Prior, 2005). Specifically, for the European context, it has also been found that public broadcasting television may be related to attitudes in a different manner than commercial broadcasting (Ariely, 2015; Holtz-Bacha, 1990; Newton, 2016; Schmitt-Beck and Wolsing, 2010). Therefore, any study investigating the relationship between television consumption and attitudes should adopt a multidimensional approach: One can expect the relationship between television consumption and public opinion to be dependent upon the specific types of television programs and stations viewers prefer (Esser and de Vreese, 2007; Hooghe, 2002; Moeller and de Vreese, 2013; Newton, 1999; Prior, 2005).

Furthermore, we know that television effects should not be situated only on the individual level. Even controlling for individual patterns of television consumption, media systems as such also have been documented to matter as they could have an indirect effect on the individual or a direct effect on the prevailing attitudinal climate within a society (Curran, Iyengar, Brink Lund, and Salovaara-Moring, 2009; Slater, Snyder, and Hayes, 2006). Crucial in this regard is that television messages can reach even those who do not watch television at all, for instance, due to the role that interpersonal interactions and network partners play in diffusing specific television content. In line with previous studies, in this article we therefore also include aggregate level measurements about market shares and global television consumption patterns (Geber, Scherer and Hefner, 2016; Schmitt-Beck and Wolsing, 2010).

In the current study, we include both individual-level and society-level information on television consumption. The use of data from the six waves of the European Social Survey (ESS, 2002–2012) allows for a comparison both over time and across countries in Europe. By including all six available waves of the ESS, we can maximize the variation with regard to market share and the presence of national minorities. We combine these population data with information on the market share of public and commercial broadcasters, an information source not yet routinely included in this kind of research, but which is based on reliable market indicators. The broadcasting systems in European countries offer enough variance to investigate our hypotheses, as some countries have a strong tradition with regard to public broadcasting, while in other countries this is not the case. Self-evidently, this kind of research design does not allow us to demonstrate any form of causality. Therefore, we include a full range of control variables that should allow us to alleviate at least some concerns about self-selection. If even after including these control variables we still find significant relations, this will lead to the assumption that it might be useful to conduct more in-depth research on this relationship.

In this paper, we first review the literature on the relationship between television consumption and anti-immigrant sentiments. We then move on to the data and methods section, and following the analysis, we conclude with some observations on both the nature of television consumption and on the societal role of public broadcasting.

Television consumption and anti-immigrant sentiments

As Western societies are becoming more ethnically diverse, issues such as immigration and diversity have shifted toward the center of public and political debate (Kriesi et al., 2012; McLaren, 2003). While the occurrence of anti-immigrant sentiments is a well-documented phenomenon, there is fierce academic debate about their origins (Ceobanu and Escandell, 2010). Scholarly attention has, among others, focused on the mass media as an important contextual determinant for the quality of the intergroup relations in society (Boomgaarden and Vliegenthart, 2009; Schemer, 2012; Schlueter and Davidov, 2013). Although the relationship between television and anti-immigrant sentiments has attracted substantial scholarly attention, studies systematically addressing the role of different types of television programs and stations in Western Europe are scarce. Moreover, studies that have considered the role of contextual patterns of television consumption in shaping opinions on immigrants are virtually non-existent (Schmitt-Beck and Wolsing, 2010). The main goal of this study, therefore, is to investigate media effects on the occurrence of anti-immigrant sentiments, by taking into account the different components of television-watching behaviour. While it has to be acknowledged that our research design does not allow causality to be detected, just establishing a relation is theoretically relevant in this regard.

The main reason to expect a relation between television and anti-immigrant sentiments is that previous research suggests that television content indeed is routinely characterized by stereotyping practices. The body of research suggests three main patterns of stereotyping: related to crime (Dixon and Linz, 2000; Iyengar and Gilliam, 2000), socio-economic status (Armstrong, Neuendorf, and Brentar, 1992; Eschholz et al., 2002; Jacobs et al., 2015; Signorielli and Kahlenberg, 2001), and religious and/or cultural issues (Das, Bushman, Bezemer, Kerkhof, and Vermeulen, 2009; Guterman, 2013). Simultaneously, we know that anti-immigrant sentiments among the population are largely based on economic and/or cultural grounds (O’Rourke and Sinnott, 2006), or prompted by fear of crime (Dinas and van Spanje, 2011). In line with these studies, we therefore expect a positive association between television time and anti-immigrant sentiments:

H1: Individual television viewing time is positively associated with anti-immigrant sentiments.

Dimensions of television effects

Back in the 1990s, researchers often contented themselves with rather general measures of television effects (Cappella and Jamieson, 1997; Gerbner, Gross, Morgan, Signorielli, and Shanahan, 2002; Newton, 1999; Robinson, 1976; Uslaner, 1998). In later studies, however, it was stressed repeatedly that specific television content matters and that therefore one should not expect uniform television effects to occur (Ariely, 2015; Esser and de Vreese, 2007; Hooghe, 2002; Moeller and de Vreese, 2013; Newton, 1999; Prior, 2005).

With regard to the type of television programs, it has been observed that a preference for programs about news, politics and current affairs is positively related to a broad range of civic attitudes and behaviors. This is related to the fact that the consumption of news results in well- informed and motivated citizens, which is a feature known to be related to civic engagement (Norris, 2000). Viewers with a preference for news media, on average, have higher levels of political knowledge (Curran et al., 2009; Prior, 2005; Vincent and Basil, 1997), higher levels of social and political trust (Lee, Cappella, and Souhwell, 2003; Schmitt-Beck and Wolsing, 2010; Shah, McLeod, and Yoon, 2001), and they participate more intensively in politics (Esser and de Vreese, 2007; Holtz-Bacha, 1990; Livingstone and Markham, 2008; Zhang and Chia, 2006). Frequent exposure to entertainment television, on the other hand, is related to attitudes and behavior which are less civic. (Besley, 2006; Eyssel, Geschke and Frindte, 2015; Hooghe, 2002). As the previous research suggests that news programs produce specific effects, our hypothesis is:

H2: Informational use of television at the individual level is negatively associated with anti-immigrant sentiments.

Earlier studies show that not only the type of program might be relevant in this regard, but also the type of broadcaster (Aarts and Semetko, 2003). Self-evidently, commercial stations will be more inclined to pay attention to entertainment programs, as they are more likely to experience pressure to keep up their market share, and as entertainment is considered commercially more attractive (Prior, 2007). In addition to this difference in program type, however, public and commercial broadcasting stations also seem to differ systematically in other, more nuanced ways. Not only are public broadcasters traditionally expected to serve the public good (Curran et al., 2009), they have often sought to promote tolerance and to report on increasing ethnic diversity and immigration in a balanced manner. Given the current salience of the debate on ethnic diversity in Europe, this would imply that they should disseminate more positive information about immigrants and ethnic minorities. Studies that have assessed whether public and commercial television depict minorities and immigrants in a diverging way, however, are scarce, although the conclusion seems to be that public television reports in a slightly more balanced manner on ethnic minorities (but see Ruhrmann, 2002; Ter Wal, 2002; Van den Bulck and Broos, 2011). An EU study has also found that public broadcasters adhere more carefully to cultural diversity in their program content (Rogers, O’Boyle, Preston, and Fehr, 2014). Furthermore, a recent study has specifically compared immigration news coverage on public and commercial stations in the Flemish context (Jacobs, Meeusen, and D’Haenens, 2016). Its conclusion was that the latter reports slightly more negatively about immigration, making more references to elements of sensationalism and tabloidization (Jacobs et al., 2016). They also found in an individual-level analysis that a preference for public news was indeed associated with lower levels of anti-immigrant sentiments compared to a preference for commercial news. This is in line with evidence stating that other types of civic attitudes and behavior are more closely related to watching public broadcasting television, and we will thus test whether this assumption holds. Public broadcasting stations are traditionally expected to play an important gatekeeping role in democratic societies as they are expected to contribute to ‘entertaining, informing and educating’ their audience (Holtz-Bacha and Norris, 2001; Newton, 2016). Most studies that have assessed the differences between public and commercial television have arrived at the conclusion that exposure to public broadcasting stations is more closely related to civic attitudes and behavior than exposure to their commercial counterparts (Aarts and Semetko, 2003; Curran et al., 2009). While the traditional public broadcasting system has come under attack in a number of European countries, in numerous societies public broadcasting corporations still have a vested mandate to inform public opinion, and the underlying policy assumption is that they do so in a specific manner (Soroka et al., 2013). One would therefore expect exposure to public broadcasting television to be negatively associated with anti-immigrant sentiments. This leads to our third hypothesis:

H3: Large market shares for public broadcasting television is negatively associated with anti-immigrant sentiments.

These effects are not just situated at the individual level, but also on a society level. Although at first sight it might seem redundant to include television time both at the individual and the aggregate level, this is the best available strategy to detect indirect media effects that operate independently from one’s own media consumption (Ariely, 2015; Slater et al., 2006). This approach is in line with the traditional two-step flow of communication theory, which assumes that leading actors within society obtain their information from the mass media and subsequently diffuse this kind of information in their personal networks (Ahn, Huckfeldt, Mayer, and Ryan, 2013). In this case, this would imply that if a large proportion of the interaction partners of an actor is exposed to a specific media content, this could still have an indirect association with the actor’s attitudes, even controlling for whether she or he is exposed to that specific content or not. This research on aggregate level effects leads to our fourth hypothesis:

H4: Aggregate television viewing time is positively associated with anti-immigrant sentiments.

In the next paragraph, we elaborate more on how these theoretical perspectives on the complex nature of television consumption patterns are relevant for investigating the relationship between television consumption and anti-immigrant sentiments.

In our empirical models we will also include various control variables that are related to self-selection, and although these do not allow us to establish causality, at least we can be more confident that the relationships we find are not spurious.

Data, method and measurements

Analysis and results

The relationship between individual and aggregate-level patterns of television consumption and anti-immigrant sentiments is analyzed via a multilevel model (Table 1). First, we present a null model with an intercept-only analysis in which the variance is split into two components: the variance between observations within societies (‘within group variance’), and the variance between societies (‘between group variance’). The size of the variances makes clear that individual-level characteristics are more important in explaining the prevalence of anti-immigrant sentiments in European societies than society-level characteristics. The intra-class correlation (ICC) shows 10.3 % of the variance in levels of anti-immigrant sentiment can be explained by society-level variables, which is substantial. This high level of intra-class correlation makes clear that multilevel analysis is indeed called for. In Model I, we add the independent variables, that is, both the television and control variables at the individual and society level, to the model.

As the sample size at the individual level is rather high (n = 162,987), we first consider the amount of explained variance by including the independent variables in Model I to reflect on the meaningfulness of the results. It should be noted that the coefficients are unstandardized since they refer to different levels of measurement and therefore cannot be compared. We can observe that in Model I the within-group variance is 3.160, and the between-group variance is 0.280, and this indicates that both on the country as well as on the individual level this model is moderately successful in improving the explained variance compared to the null-model. Therefore, it is warranted to move on to a meaningful interpretation of the coefficients. With regard to the individual-level television variables, the results of Model I make clear that H1 and H2 are supported by the analysis: On the individual level, there is a strong positive relation between television time and anti-immigrant sentiments (H1), while frequent exposure to news and programs on politics and current affairs is negatively associated with anti-immigrant sentiments (H2), and this effect remains robust throughout the further analysis. If we consider the television variables at the aggregate level, we find in this model that H3 is confirmed: There is a positive effect of the market share of public broadcasting. However, if in Model II we subsequently add the aggregate level of viewing time, this variable helps us to explain the variance (confirming H4), while the effect of the market share of public broadcasting is rendered not-significant. Hence, while at first sight, a larger market share of public broadcasting seems to have an independent effect, in the subsequent analysis this effect is completely taken up by the fact that apparently in these countries, the average television viewing time in society is also higher. In particular, the fact that in some societies the average viewing time is much higher therefore seems to be particularly relevant. Still, given the large sample size, the significance levels should be interpreted with caution.

Model II also includes control variables at the individual and aggregate level, and here the results mostly confirm previous research (Ceobanu and Escandell, 2010). Older respondents have higher levels of anti-immigrant sentiment, and so have respondents that have recently been victimized as a result of burglary or an assault. Respondents that are less politically interested hold higher levels of anti-immigrant sentiment as well. But particularly respondents’ educational level, income satisfaction and life satisfaction are negatively related to anti-immigrant sentiments. Considering the control variables at the aggregate level, we do not find an effect of the GDP level or the inflow of foreigners. It is clear, therefore, that individual-level predictors are decisive in explaining anti-immigrant sentiments. However, this does not alter the fact that, even under control for these individual characteristics, patterns of television consumption level continue to make a significant contribution.

Discussion

In this article we investigated the relationship between exposure to specific types of television program and broadcaster and the prevalence of anti-immigrant sentiments in society. We observed that both individual and aggregate patterns of television consumption are associated with attitudes towards immigrants. At the individual level, heavy television viewers are more likely to hold negative attitudes towards immigrants, although frequent exposure to information programs counterbalances these attitudes. This confirms the notion that the effect of television is strongly dependent on the type of program that is being watched, and our findings add to the rather pessimistic view of the expansion of entertainment television. Despite some studies indicating that stereotyping also occurs in news broadcasts, a preference for news broadcasts still seems to be negatively related to anti-immigrant sentiments.

At the society level, it seems that in societies where on average more time is being spent on watching television, there are higher levels of anti-immigrant sentiments. It is important to repeat here that this relationship is robust, even after controlling for the time the individual respondent spends watching television. The notion that television also has an aggregate-level effect, extending toward those that do not watch television or only now and then, therefore is clearly supported. In societies where the average time is higher, everyone seems to be affected, even controlling for the respondent’s individual viewing habits.

This study thus provides support for the view that when studying the relationship between television consumption and attitudes, one should adopt a multidimensional perspective that distinguishes between specific types of television program and broadcaster. In this way, our study corroborates previous findings that news viewing habits and the amount of time watching television are closely related to civic attitudes. While this relationship was already demonstrated for political knowledge, social and political trust and political participation, our study is the first to systematically address the relationship between differential television consumption habits and anti-immigrant attitudes. For this kind of negative attitude, too, we do find a negative relationship with a preference for news programs. Despite the fact that previous research has highlighted the presence of stereotyping in broadcast news, apparently following the news and being informed about society still reduces anti-immigrant sentiments. To some extent this might be related to higher levels of political interest and knowledge among those who watch the news (although we controlled for respondents’ educational level in the analysis and their socioeconomic status and political interest), but there is also clearly no observable relation with stereotyping in news broadcasts. News stations may provide more background information in their broadcasts, and learning about immigrants could provide news recipients with individuating information about these social groups. Literature on stereotypes suggests that individuating information has the potential to erode categorical or stereotypical judgments of social groups.

The explanation for the associations we have found between television consumption and anti-immigrant sentiments, however, is not evident. First, it should be mentioned that our study relies on a reported preference and reported habits with regard to watching television. This kind of survey question most likely leads to social desirability, and therefore it would be dangerous to base causal inference purely on this survey question. We have to take into account the possibility that the same motives that drive respondents to report their preferences for information media too positively, might also lead to a biased answering pattern with regard to anti-immigrant sentiments. Content analysis, therefore, should demonstrate more systematically whether there is a difference between various programs and broadcasters in the way they depict ethnic and cultural diversity in society. A recently published study suggests that in Flanders this is the case: Public broadcasters report in a slightly more positive manner about immigration and ethnic diversity compared to commercial broadcasters, by focusing more on positive advantages of diversity for society (Jacobs et al., 2016). It is important, however, to consider that in countries that are heavily dominated by commercial broadcasting, there is a strong incentive to stimulate the overall level of time that is being spent on watching television. The longer, on average, people watch television, the more potential there is for gathering revenue. The same kind of market logic, however, does not apply to systems that are dominated by public broadcasting, as most public broadcasters are not dependent on revenue for advertisements. Another approach which would strengthen any causal claim in this regard is the adoption of a research design which allows for a direct test of causality, such as a longitudinal or time-series design. Data limitations prevent us from applying such a design, as this kind of data is currently not available, and in future research experimental designs should be envisioned to determine whether the associations we observed in this study might also be explained in a more causal manner.