Agenda setting and the political economy of fear: How crime news influences voters’ beliefs

Abstract

In this article, we analyse the mechanisms of agenda setting by focusing on the determinants of individual attitudes towards crime and investigating the role played by the media. After a brief literature review supporting the relevance of the selected topic of inquiry and the presentation of our analytical framework, we study the persuasion effects of mass media. More specifically, we investigate how TV exposure can shape individual perceptions of specific issues such as crime, and then focus on the effects of exposure to crime news on voting decisions. Using the Italian 2001 general election as an important case study of TV power concentration, we provide evidence that media-induced agenda setting enhanced the salience of the crime issue in voters’ minds during the 2001 Italian general election and contributed to the victory of the coalition led by Silvio Berlusconi. Interestingly, our results are partially driven by the switch of previous left-wing voters to voting for the centre right because of exposure to crime news.

Keywords

Fear media systems electoral behaviour Italy

Introduction

The role of fear in politics is a hotly debated issue in contemporary political science and political sociology literature. A number of scholars have argued that geopolitics is now governed by emotions and that fear governs Western politics in both international and national contexts: ‘fear leads to a defensive attitude that reveals and reflects the identity and the fragility of a person, a culture, or a civilization at a given moment’ (Moisi, 2009: 92; see also Furedi, 2005). Furthermore, a growing body of literature is focusing on the relevance of domestic factors, such as economic conditions, immigration and crime in driving the emotional and fearful reactions of citizens and voters. In such a context, the role of media in shaping human behaviour and ultimately the outcome of political elections is crucial in understanding the way in which political elections may be dominated by emotions.

In this article, we consider critically these arguments and attempt to study the political influence of TV through a specific channel: fear of crime. The definition of fear we adopt is that of Witte (1992): ‘fear is a negatively-valanced emotion, accompanied by a high-level of arousal, and elicited by a threat that is perceived to be significant and personally relevant’. In our contribution we try to evaluate the impact of TV media on agenda-setting, for example, on the selection of the key issues of political agendas. More specifically, we study a context in which a medium (TV) transmits a disproportionate amount of crime news, and crime is a relevant topic in the political agenda. Consequently, in our main research hypothesis, voters – scared by crime – vote for the party that they think can best address the issue.

In terms of policy fields, we have selected crime as a policy area since – after economic stability – it is today considered the most relevant among the citizens surveyed through World Values Survey (WVS, 2005–2008). Furthermore, due to the exploratory nature of the article, we will focus on one crucial case (Gerring, 2007). Italy is particularly interesting due to its current limited media (especially TV) pluralism: it has been defined as ‘partly free’ in the most recent annual Freedom House survey (2011), ranking 32nd, one of the three European Union countries (together with Bulgaria and Romania) not to be considered as ‘free’. Put differently, we are mainly interested in verifying the impact of crime news coverage by TV news programmes in Italy in the period of the 2001 political election on the salience attributed by voters to crime and, through this channel, assess the influence TV media exerts in a given media system. In order to answer our research question, we use microdata from the 2001 Italian Electoral Survey and propose a methodology that shows how crime news affects the beliefs of voters and their electoral behaviour.

The main findings of the article can be summarized as follows: in a context of direct TV power concentration in the hands of key political leaders, the capacity to shape the political agenda and impose it on the media agenda is particularly high. In our focus on a specific electoral period, we provide robust evidence of such a capacity with respect to reported crime, which was declining in reality but due to its electoral salience, increasingly covered by TV news on the channels owned by the political leader and his family. The article develops the arguments by introducing the analytical framework and the overall research design (section 2), presenting the methodology and the data (section 3), the results (section 4) and concluding remarks (section 5).

Political agenda setting, media agenda setting and impact analysis: Research design and hypothesis

Media and TV models

The role of mass media in political marketing is crucial in electoral strategy to gain or build consensus around policies and politicians. According to the growing literature on media capture, newspapers provide press coverage of the issues that appear to be politically more important (Larcinese et al., 2011); media are also relevant in framing the political agenda (McCombs, 2005; McCombs and Shaw, 1972) and the agenda setting dynamics may vary since they are grounded in different media systems (Hallin and Mancini, 2004). With this regard, the most important contribution to the literature on media regulation systems is Hallin and Mancini (2004). The authors suggest looking at media regulation with respect to four dimensions, and according to their analysis, three ‘media systems’ are identified in Europe: (a) the polarized pluralist model (among other countries France and Italy) which is characterized by low levels of newspaper circulation and journalistic professionalization as well as by high levels of political parallelism and state intervention; (b) the democratic corporatist model (among other countries Germany and Sweden) which is characterized by high levels of mass circulation press and political parallelism, a high degree of journalistic professionalization and state intervention; and (c) the liberal model (among other countries United Kingdom and United States) which is characterized by medium levels of media circulation, strong professionalization and low levels of parallelism – with the exception of the United Kingdom – and state intervention. Although the authors stress that the three models are ‘ideal types’ and that media systems are not static, their main argument is that variations in political systems will persist and continue to shape (and be shaped) by the media (Puglisi, 2011).

Although very promising and widely cited, Hallin and Mancini’s contribution does not fully take into account the specific ‘agenda setting’ potential of TV news – which, as it is well known, has become increasingly relevant as a source of political information in contemporary democracies. Therefore, if we focus on one specific medium – TV – and we consider the degree of TV power as a classifying criterion, then we may question whether Italy remains an example of a polarized pluralist model. Italy is a crucial case since – beyond public broadcasters – for over a decade the overall power concentration of the leader of the centre right, Silvio Berlusconi and his family has been extremely high (see also Durante and Knight, 2012); furthermore, together with the main private TV channels owned by the Berlusconi family (Canale 5, Retequattro and Italia 1), previously to the years under scrutiny in this article all the public channels (RAI 1, RAI 2 and RAI 3) have been controlled by various Parliamentary nominees (Mancini, 2009), among which we could also find people nominated by Berlusconi’s centre-right representatives. Therefore, although pluralism can be seen in the appointment of the key public TV managers (Mancini, 2009: 125–126) and Berlusconi’s party was in opposition during the 1996–2001 period, the fact that three out of the six main TV channels have been constantly fully controlled by the leader of the centre-right coalition is per se a clear indicator of how concentrated the Italian TV system was. We hypothesize that this TV power concentration may have significant effects in defining the political agenda, especially in times of elections; put differently, taking into account the public and private TV media system, over the past twenty years (1996–2016) the Italian model is less pluralist and more heavily concentrated, being better labelled as an example of a highly concentrated model rather than a polarized pluralism model.

TV and agenda setting in a highly concentrated model

For several years, under the same ‘agenda setting’ label two sets of literature have lived together. From a policy analysis perspective, agenda setting has been defined by Kingdon as ‘the list of subjects or problems to which government officials, and people outside of government closely associated with those officials, are paying some serious attention at any given time…the agenda setting process narrows [a] set of conceivable subjects to the set that actually becomes the focus of attention’ (Kingdon, 1995: 3). The main focus of the path breaking book by Kingdon and the empirical research carried out by other policy analysts has been the moment in which a given social problem gains public attention and triggers a policy cycle (Howlett et al., 2009). Kingdon’s analysis remains still unrivalled when it comes to macro-determinants, but if we want to focus on the specific role played by the media in agenda setting then we are somehow ‘lost in operationalization’ since within the ‘grand theory’ provided by Kingdon not much space is left for ‘micro’ variables such as the media – and TV more specifically.

To be sure, a relevant media research literature on agenda setting has emerged and consolidated over the years by further trying to differentiate agenda setting from other concepts such as framing and priming (for a succinct overview, see Weaver, 2007). More generally, since the 1940s, a number of studies on the impact of the media on voters’ preferences have been conducted (among others: Campbell et al., 1960; Converse, 2006; Lazarsfeld et al., 1944), but only in the past two decades has a more nuanced analysis of the role of the various media been provided (see for example Brians and Wattenberg, 1996). Still partially overlooked is the role played by the media with respect to selected issues during the electoral period (for some exceptions, see Petrocik, 1996; Wanta and Ghanem, 2007).

A third strand of the literature we consider focused on the impact of TV diffusion on social behaviour (see amongst others Gentzkov and Shapiro, 2004; La Ferrara et al., 2008) and on voting behaviour (Della Vigna and Kaplan, 2007; Enikolopov et al., 2011; Lawson and MacCann, 2004; White et al., 2005).

In order to better grasp the relationship between media power and agenda setting capacities, we define: (a) political agenda setting as a process of making a social problem or issue politically salient, for example, object of debate among institutional (governments, legislatures, etc.) and non-institutional actors (pressure groups, parties, social movement, etc.); and (b) media agenda setting as the process of making a social issue mediatically salient, that is, object of debate within the media system of a given political system. A clear distinction between the two types of agenda setting sheds specific light on the mechanisms through which media power may foster political agenda-setting capacities and therefore a series of more specific hypotheses can be derived and tested empirically. To be sure, empirically there may be overlaps between political agenda setting and media agenda setting, but analytically speaking it seems particularly appropriate to distinguish between these two types of agenda since – especially during the electoral period –political agenda setting may be particularly influenced by media agenda setting, especially when the issues covered by the news may be politically salient in terms of acquiring further consensus.

In our study, we focus on the moment where the links between the two above mentioned forms of agenda setting are at their peak: the electoral period. In fact, during the electoral period, political agenda setting is heavily driven by political campaigning and the contraposition of electoral programmes – which are contrasting sets of political agendas over which the electorate is called to vote. Furthermore, the electoral period is also a moment when the main focus of media news is the political competition among the various candidates – whatever the electoral system may be. Therefore, during the electoral period – including both the electoral campaign and the day(s) when votes are cast – it is accurate to state that the political agenda setting and media agenda setting are focusing on the same objects and the links between the two types of agenda setting are particularly strong. Finally, during the electoral period of particular relevance is the set of leading actors who intervene in both agenda setting types. In cases of extensive TV pluralism (low concentration model), we assume the key actors are different and no key actor is in significant control of both agenda setting processes; in case of limited media pluralism (high concentration model – the Italian case), we assume that the actors may be less differentiated or at least that some hierarchy may be established – typically under the form of a political control over the media. This is not to say that the Italian case during the electoral period we analyse (2001) was an example of no media pluralism: as Berlusconi was not Prime Minister, he did not control both public and private TV. Nevertheless, he clearly had significant control with respect to private television and this had evident repercussions on how differently crime news was reported by private and public television (see Mancini, 2003).

By focusing on one medium (TV) the analytical framework we propose here is one in which political leaders, through the media, set the political agenda, - the media transmit news strictly related to the issue (for example, crime) which is particularly crucial for voters; and, voters receive this signal and form, through fear, beliefs and political attitudes regarding the salience of the issues (Figure 1; see Supplementary Material at http://ips.sagepub.com). Although we will return to this observation in detail in the next section, the point of departure of our analysis is that the more powerful a political leader is in terms of media (TV) power concentration, the more probable it is that the political agenda setting will influence the media agenda setting with significant repercussions on the perceived salience of a given policy issue (crime) and related voting behaviour.

Political agenda setting can be understood by looking at the preferences expressed by relevant political parties on given policy issues, whereas media agenda setting can be measured via the coverage provided with respect to a certain policy issue.

Before going any further, however, we still need to specify the main features of the selected issue – crime – and its social consequence – fear – and connect them to our research framework.

The policy issue: The political economy of fear

When do politicians and media have an incentive to scare the voters? Lupia and Menning (2009) study the case in which politicians use fear to scare the voters and gain support for policies. They argue that many researchers tend to define exogenously (ex-ante) the issues that are supposed to scare the voters, hence the authors focus on the sensitivity of voters to particular items, which are the ones that politicians use to build consensus over given policies. In other words, they argue that fear is a useful political tool only if limited to certain issues, defined by the sensitivity of voters.

Voters often have superficial beliefs about policies (mainly economic policies), and politicians reinforce those beliefs through propaganda in order to increase the probability of being re-elected (Canes-Wrone et al., 2001; Caplan, 2007). Individuals do not necessarily vote on the basis of political programmes or beliefs, rather they may cast their vote focusing on performance evaluation of politicians or party identification. In general, media and politicians know the value of fear and use it to increase public support and audience (Gardner, 2009). Politicians may gain by promoting threats, with the idea that they could be the solution to the problem. This implies that part of the political campaign may be based on shaping and using emotions, especially when focusing on less educated voters (Brader, 2006). Yet, it is often the more informed who are more influenced by fear-driven news and advertising; hence, the role of media – and especially of TV – is crucial for spreading fear. In fact, Gardner (2009) reports that broadcast media give disproportionate coverage to criminal and violent events, up to 10–30% of newspaper content. This is of vital importance for the aim of our research as we hypothesize that media (TV) influence voters’ beliefs not only through direct exposure of politicians, but also through the disproportionate reporting of news correlated with issues that are high on the political agenda of (one of the) political coalitions.

In order to try to link the various strands of the literature mentioned in the previous subsection, we first provide some conceptual clarifications and then move on to a formalized analysis of the impact of TV on the overall agenda setting in an exploratory case – Italy in 2001 (when a mainly majoritarian electoral system was in place).

Our empirical analysis focuses on the 2001 Italian national elections for two main reasons:

the governance structure of the TV system is concentrated in the hands of the parliament and of one large private firm owned by the leader of the centre-right coalition, and his family;

in the spirit of Lupia and Menning (2009) we focus on a specific issue that is unanimously considered to be politically salient for citizens/voters - crime. The electoral campaign of 2001 placed considerable emphasis on this issue (ITANES, 2001).

Legnante and Baldassarri (2010) suggest that the media became crucially important in Italy after 1994. In the 2001 electoral campaign, the centre-right coalition was particularly successful in communicating the salience of crucial issues in the programme, such as crime, immigration and taxes (Legnante and Baldassarri, 2010).

A first assessment of the diverging pattern between crime news and reported crime rates can be conducted by comparing the growth rates in the minutes of crime coverage on TV and the growth in reported crimes. As Figure 2 (see Supplementary Material at http://ips.sagepub.com) shows, news programmes on private channels (TG4, TG5 and Studio Aperto) have considerably increased crime news coverage, whereas public channels have decreased coverage. It should also be noted that between 2000 and 2001, the growth rate in the number of homicides was negative and equal to −5.61%. Similarly, the growth in the number of overall reported crimes was negative, accounting for −1.97% (Ministero dell’Interno, 2006).

The picture that emerges from Figure 2 is that private channels increased disproportionately the coverage of crime news and this, according to our view, could be due to an alignment to the issue proposed in the political programme of the centre right. At this point, a distinction must be made clear. An increasing body of literature is addressing the issue of the relationship between media and politics, focusing mainly on measurement issues. In this paper, we assume this link and mainly observe the political consequences of crime news exposure. Note that this assumption is corroborated by the characteristics of the Italian media system outlined previously.

In sum, our main research hypotheses are the following: during the electoral period, in cases of power concentration in the TV system, the (TV) media agenda will be heavily influenced by the political agenda (RH1) and media information will be selected to correspond to key policy concerns of specific segments of electorate (RH2) – even when the ‘hard’ facts do not correspond to the channeled information.

Methodology and data

In terms of operationalization, we look at political agenda by focusing on the relevance on crime of the party manifestos of the centre right and centre left coalitions – Casa della Libertà (House of Freedom) and L’Ulivo (Olive Tree), respectively. The available evidence suggests that the centre right coalitions prioritized crime issues to a greater extent than did the centre left. Of the 15 pages which articulate the so called Contratto con gli Italiani (Contract with the Italians), which was the main document used by the leader of the centre-right coalition during the electoral period, crime issues are the most important topic among the five ‘contractual points’: almost half of the document is devoted to such issues (Berlusconi, 2001). In the case of the centre left coalition, which presented a more traditional electoral programme of 117 pages, crime issues are much more marginal with respect to other issues: only three pages are dedicated to crime issues, and they are not directly mentioned in the key six challenges outlined in the manifesto (L’Ulivo con Rutelli – insieme per l’Italia, 2001: 15). In addition, secondary research has emphasized the greater attention paid by the centre-right coalition towards crime issues than the centre left (Mancini, 2003; Marletti, 2002; Newell, 2003).

In order to test the hypotheses presented in the previous section, we take an innovative route as we aim to estimate the marginal impact of a particular type of news (e.g. crime news) on voters’ beliefs and on their voting behaviour in the 2001 election. For this purpose, we propose a model in two steps: in the first step we estimate the impact of crime news coverage by major news programmes on the subjective importance of crime as an issue to be addressed by the new government; and in the second step, we estimate the impact of the predicted importance of crime on voting behaviour.

Put differently, we estimate a model aimed at explaining voters’ choice as a function of the salience they attribute to crime, which is linked to measures of crime news watched on TV and an index of diffused crime in the region of residence of the voter.

It should be noted that this approach allows us to test the framework we have proposed in section 2 since in the first step we estimate an equation in which, given the quantity of crime news potentially watched on the TV, voters form their beliefs about the salience of the crime issue, whereas in the second step those beliefs are the bases upon which voters elaborate their electoral preferences.

Before describing our empirical analysis in detail, it is useful to mention clearly that we do not model explicitly (i.e. in statistical terms) the relationship between the political system and the medium (TV). This choice is driven by the governance structure of the Italian TV market and seems to be a reasonable assumption, as discussed in the previous section.

More formally, in the first step we estimate the following equation for voter i living in region j watching mostly news programme k and then news programme h

C r i m e_i m p_{i j} = α + β X_{i} + γ C r i m e_{j} + ϕ C r i m e_n e w s_{k h} + φ F i r s t N P_{i} + θ S e c o n d N P_{i} + ε_{i}

(1)

where Crime_imp is the subjective importance of crime in the government action, X is a vector of individual characteristics, Crime is a crime index in the region of residence of the voter, Crime_news is an index aggregating crime news coverage of the most watched and of the second most watched news programmes, FirstNP and SecondNP are fixed effects for the most watched and the second most watched news programmes respectively. The inclusion of such fixed effect is thought to capture all other news programme-specific characteristics (eventually not observed) other than the amount of crime news which may influence the perception of the relevance of crime.

The rationale for the functional form in (1) is that individuals may shape their beliefs on the basis of both the actual level of crime in their environment (Crime) and on the exposure to crime news (Crime_news).

In the second step we estimated the following equation

V o t e_{i} = a + b X_{i} + c C r i m e_i m p_{i} + ϖ_{i}

(2)

where Vote is an indicator of voter choice in 2001 political elections, that is, vote for the centre-right coalition and Crime_imp is the predicted value of Crime_imp as from equation (1), X is a vector of individual and regional controls, as in equation (1).

Our data are mainly from the 2001 Italian National Electoral Survey (ITANES), carried out by the Istituto Cattaneo, merged with other data from different sources.

The importance of crime for voter i is elicited by asking the respondent to state a value on a scale between 1 and 4, where 1 means ‘It is not important at all’ and 4 means ‘It is very important’, to the question ‘I will now list some social and economic problems discussed during the recent election campaign. According to you, how important is each of them in the case of Italy?’ One of the items was crime. Hence, each individual was asked to state his/her perception of the importance of crime on a four-point scale. One of the main problems with this measure of crime salience is that people with similar demographic, economic and social characteristics can use different subjective scales to evaluate the same phenomenon or to respond to the same question (Holland and Waimer, 1993). In order to overcome this problem and to increase the comparability across individuals, we have firstly computed the mean of the importance attributed to the fifteen items asked in the question, hence we have computed the average importance for voter i, $M e a n_{i} = \frac{\sum_{l = 1}^{15} I_{l}}{15}$ where I_l is the importance attributed by individual i to item l. A similar procedure was carried out for the computation of the standard deviation of the grades attributed to the fifteen items. Hence, our measure of (relative) crime importance is the grade given to crime by individual i standardized with respect to answers given as for other items by the same respondent

C r i m e_i m p_{i} = \frac{I_{C r i m e, i} - M e a n_{i}}{S t d D e v_{i}}

(3)

We used different measures to identify the impact of crime in the region of residence of the respondent. In particular, we made use of a diffused crime index and of an index of violent crime. Since the National Electoral Survey was conducted in May–June 2001, we consider both year 2000 and 2001. Both variables are at regional level and the source is ISTAT, Statistiche giudiziarie. In what follows we will report only results for the diffused crime index in 2001, as it provides a better fit of the model and is a better instrument for Crime_imp.

Crime news coverage, in terms of minutes, is from Marchese and Milazzo (2001). In Italy, six major news programmes cover almost all the audience, they are: TG1; TG2; TG3; TG4; TG5; and Studio Aperto. The former three are on State-owned channels, while the latter three are on Mediaset channels owned by companies where the majority of shares are in the hands of Silvio Berlusconi’s family. In order to increase the variance of our ‘crime news coverage’ variable, we will exploit the information retrieved by answers to the question: ‘Which news programme do you usually watch most? (If more than one, please tell me the most watched and then the next)’. Hence, for most of respondents in our sample, we have the most watched and the second most watched news programme. This allows us to carry out a principal component analysis across individuals for the crime news coverage of the two news programmes most watched by voters. In this way we exploit all possible information at individual level and increase the variability. Note that this variable cannot be interpreted as individual exposure to crime news; rather, is it meant to indicate the propensity of a particular mix of news programmes to spread crime news.

To sum up, we estimated an instrumental variable probit model, where the instruments for crime salience are both crime news coverage of the most watched news programmes and the presence of crime in the region of residence of the respondent.

In order to identify the parameter $φ$ in equation (1), we need to control for fixed effects of the news programmes (both the first and the second most watched ones). However, the choice of the TV programme could be endogenous to our model. In order to control for this case, we estimated a sort of zero-th stage in which the choice of the news programmes is endogenized by estimating a function of the type

F i r s t N P_{i} = A + B C o n t r o l s + C F a t h e r + D A g e d i s t + ξ_{i}

(4)

where Father is a vector of characteristics of respondent’s father and Agedist is the Euclidean distance between individual age and the age of the editor of the news programme. Controls include father’s education, political preferences and interest when the respondent was 14 years old. The rationale for including the father’s characteristics is to proxy prior beliefs of the individual in choosing news programmes, while the Euclidean distance allows for inclusion of an exogenous variation in the choice of the news programme.

Equation (4) is estimated across individuals for both the most watched and the second most watched news programme by means of a multinomial probit model. A similar equation has been estimated for SecondNP. It should be noted that this procedure does not fully avoid problems of endogeneity, although it is considerably lowered. In fact, our assumption is that the supply of crime news is exogenous to the individual characteristics, hence the fixed value for news programmes are necessary to identify the marginal impact of exposure to such news on the salience of crime as perceived and stated by the individual.

In other words, we assume that once the choice of a given mix of news programmes is made, the single voter has no influence on the decision to supply a given level of crime news made by the editorial board of the news programme. It should be noted, however, that the fact that we omit ideological variables from the models may be relevant as ideology may significantly influence the choice of news programmes made by individuals. Unfortunately, we cannot explicitly test for this hypothesis and we think that this might be a relevant limitation of our model.

Furthermore, parameter $φ$ is identified only if crime news is not collinear to any omitted variable of the news programme mix. This condition is mildly satisfied by including in (1) (conditional) probability to watch the six news programmes’ fixed effects and by considering a hybrid measure of news coverage, as given by the linear combination of crime news coverage of the most watched and the second most watched news programmes whose weights are given by a principal component analysis. This aggregation procedure makes the likelihood of the existence of an omitted variable bias less important. However, in what follows we will report also results when including a different type of news, often considered highly related to crime, that is, the relevance of immigration news.

Finally, our first measure of voter’s choice, Vote, is the probability to vote for the centre-right coalition, hence for Berlusconi. However, there is an extensive literature on the flexibility of swing voters’ preferences (Feddersen and Pessendorfer, 1996; Robinson and Torvik, 2008). Given this characteristic, it is very likely that media may exert their influence especially on those individuals, hence by influencing their decision to swing their vote. In the present article, we do not characterize those voters on the basis of self-reported ideology on a standard ten-point scale, because of weak comparability across individuals. Rather, we focus on the probability to swing the vote from the centre-left coalition in the 1996 election to the centre-right coalition in the 2001 election, and consider the crime policy issue since it is one of the most relevant for both centre right and centre left electorates (ITANES, 2001; Legnante and Baldassarri, 2010). As mentioned above, our main source of data is the 2001 Italian National Electoral Survey conducted by Istituto Cattaneo in the aftermath of the 2001 election, the period 18 May–18 June 2001. In this survey, about 2,700 individuals were asked to state their vote for both the Senate and the Low Chamber, as well as to respond to a long series of questions regarding their socio-demographic characteristics and their values or political opinion. In the present article, we decided to focus on voting behaviour for the Lower Chamber (the majoritarian part) which is also often studied as a proxy for as the most representative of the Italian legislature since voting age is 18 (whereas for the Senate the voting age is 25).

As shown in Table 1 (see Supplementary Material at http://ips.sagepub.com), the survey slightly over-sampled votes for the centre-right coalition, which, in the case of our analysis, especially for swing voters, is not particularly problematic as all our regressions are weighted with weights provided by ITANES. Table 2 (see Supplementary Material at http://ips.sagepub.com) reports summary statistics of the main variables of interest. In Table 3 (see Supplementary Material at http://ips.sagepub.com) we report both crime news coverage by news programmes in the first semester of 2001 and their relative audience, according to the 2001 ITANES survey. News programmes on State-owned channels (TG1, TG2 and TG3) casted 921 minutes with 54.64% of respondents reporting one of them as the most watched news programme, while Mediaset channels (TG4, TG5 and Studio Aperto) had 1.099 minutes of crime news with 45.36% of reported audience¹.

Discussion

As stated previously, in the 2001 electoral campaign, the centre-right leader, Silvio Berlusconi, signed the so-called ‘Contratto con gli italiani’, in which the fight against crime in the cities was the only non-economic item in a list of five issues to be addressed by the new government. Furthermore, the most relevant issues debated during the electoral period were crime, immigration and taxes, with crime as the major shared policy issue between centre-left and centre-right voters (ITANES, 2001). To corroborate our assumption that exposure to crime news leads individuals to vote for the right, in Table 4 (see Supplementary Material at http://ips.sagepub.com) we report regressions in which the dependent variable takes the value of 1 if the respondent has answered ‘centre right’ to the question ‘I would like to know what kind of government is more able to deal with the following problems: a centre left or a centre right government?’ and zero otherwise. In models 1 and 2 we do not consider potential endogeneity in the choice of the news programmes and consider both the sum and the principal component of crime news coverage by the first and second most watched news programmes. We find that our second measure of crime news coverage is highly significant and related to the support for the centre right for fighting crime. In models 3 and 4 we instrument the choice of the news programmes with equations in the form of (4). Interestingly, it emerges that all the fixed effects for the first most watched news programme lose their significance, whereas the second ones are now significant, and even more importantly, both measures of crime news coverage are now significant and positive. In addition, to be noted is the fact that the presence of crime in the region of residence of the respondent is not significant.

Hence, results in Table 4 seem to corroborate our assumption that crime is a particularly salient issue for the centre-right electorate.

Our instruments for the identification of the impact of crime salience on voting behaviour are the diffused crime index and crime news coverage (both sum and principal component). In order to be suitable instruments for our analysis, both variables should be relevant in explaining voting behaviour and not correlated with omitted variables in the second stage regression. In Table 5 (see Supplementary Material at http://ips.sagepub.com), we report results of reduced form equations, from which it emerges that both variables are strongly correlated with the choice to vote for the right. In addition, to be noted is that crime news fixed effects preserve their sign across the specifications, although, when considering possible endogeneity, their significance is lower.

Table 6 (see Supplementary Material at http://ips.sagepub.com) reports the estimation of our instrumental variable probit model. In Panel A we report the estimates of our first step estimation. It turns out that, in models 1 and 2 both the sum and the principal component measures of crime news coverage are significant and positively correlated with the standardized measure of crime salience. Furthermore, the index of diffused crime is positive and significant. Second stage regression results are in Panel B and confirm our hypothesis that crime salience is a major source of votes for the right.

In models (3)–(6) we split the sample between centre-left voters and centre-right voters in 1996. It should also be noted that the number of observations drops because of individuals not remembering or not stating their previous vote. Interestingly, it emerges that crime news coverage is an important determinant of crime salience only for leftist voters in 1996. Hence, it seems that persuasion in the form of crime news is very effective only for swing voters. The p-value of the Hansen over-identification test is always very high.

As stated above, answers to the question on the salience of crime may be difficult to deal with because of different psychological scales of the respondents. We have made use of a standardized measure of crime importance as defined by (3) to ensure comparability across individuals. This new measure has the appealing characteristic to be continuous; however, it does not exclude per se the possibility that different individuals rank crime with respect to other political issues differently. In order to exclude also this last source of bias, we have considered answers to the question ‘Among the problems previously listed, you considered the following very important. Which one of them do you consider the most important? And which would be the next most important?’ Hence, we have coded a dichotomous variable, which takes the value 1 if the respondent has reported crime to be the first or the second most important problem to be addressed by the Government. Table 7 (see Supplementary Material at http://ips.sagepub.com) reports the results of this robustness check and shows the robustness of our results with respect to this different specification.

Tables 8 and 9 (see Supplementary Material at http://ips.sagepub.com) report further sensitivity analysis of our model; in particular, in Table 8 we add in the first stage a further item, that is, coverage of immigration news. Interestingly, this variable is never found to be significant in the first stage regressions across the specification. Hence, it seems that our standardized measure of crime salience is reactive to crime news coverage, but not to immigration news. Finally, in Table 9, we consider crime news four months before the election, in order to have a better identification. Interestingly, all the coefficients are larger than the ones in the specifications in Table 6 and the results are confirmed.

Finally, although the results of the Hansen-J test for over-identifying restrictions give comfortable values about the exogeneity of our proposed items, in Table 10 (see in Supplementary Material at http://ips.sagepub.com), we present results of some placebo regressions in which crime news coverage and crime salience are regressed on the decision to vote for the centre-right coalition. Also in this case our assumption about the exogeneity of crime news coverage seems to be verified since the dummy variable indicating whether voters actually voted for the centre right is not significant, as is also the case for model 4 where the dependent variable is crime salience and where we control for the other items. These tests confirm our results.

In sum, from our data, it seems that TV may influence voters’ perceptions through specific channels. In particular, we have proposed sound evidence that an increase in crime news coverage in major TV programmes increases the salience voters assign to crime (also by taking constant crime rates) and hence the probability to vote for a specific party coalition.

Conclusion

Indisputably, media pluralism is one of the key features of contemporary democracies. In this article, we have focused on the consequences of media (TV) power concentration on voting behaviour. We have used Italy as a crucial case due to its highly concentrated TV system, and crime as a policy issue due to its overall salience for the electorate studied. First, we have characterized the Italian TV system as ‘highly concentrated’, in contrast to previous, more general analyses of the Italian case as a polarized pluralist system, and we have theoretically distinguished political agenda setting from media agenda setting. Second, we have shown how the political agenda has influenced the (TV) media agenda, especially with reference to the TV channels directly connected to the centre-right political leader. Third, we have shown how TV information has been selected in relation to key policy concerns of specific segments of the electorate – either centre-right voters or swing centre-left voters. We have then presented robust evidence on the capacity of the political agenda – framed in particular by the political leaders who have greater control of TV – to shape the media agenda and hence distort voters’ beliefs. With specific reference to Italy and to crime news, we have found that the distortionary effect of TV was a substantial explanatory factor of the Italian 2001 election outcomes. Beyond the Italian case, our contribution highlights the electoral consequences of limited TV pluralism in a context where elections are highly competitive and swing voters are particularly crucial for winning.

Footnotes

Acknowledgements

We would like to thank Luca Corazzini, Valentino Larcinese, Riccardo Puglisi and three anonymous referees for comments on various versions of this article.

Funding

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Notes

Author biographies

Paolo R Graziano is Professor of Political Science and Policy Analysis at the University of Padua and Research Associate at the European Social Observatory, Brussels. His work has appeared in political science and public policy journals such as: European Journal for Political Research, Governance, Journal of European Social Policy, and The Annals of the American Academy of Political and Social Science.

Marco Percoco is an Assistant Professor of Regional and Transport Economics in the Department of Policy Analysis and Public Management at Università Bocconi, where he is also affiliated to Dondena Center and CERTeT. His main research interests are in the evaluation of regional and transport policies and in historical economic geography.

References

Berlusconi

Silvio

(2001) Contratto con gli italiani. mimeo. Available at: http://www.tgcom24.mediaset.it/documenti/contratto.pdf

Brader

Ted

(2006) Campaigning for Hearts and Minds. Chicago, IL: University of Chicago Press.

Brians

Craig Leonard

Wattenberg

Martin P

(1996) Campaign Issue Knowledge and Salience: comparing reception from TV commercials, TV news, and newspapers. American Journal of Political Science 40(1): 172–193.

Campbell

Angus

Converse

Philip E

Miller

Warren E

Stokes

Donald E

(1960) The American Voter. New York, NY: John Wiley and Sons.

Caplan

Brian

(2007) The Myth of the Rational Voter: Why Democracies Choose Bad Policies. Princeton, NJ: Princeton University Press.

Canes-Wrone

Brandice

Herron

Michael C

Shotts

Kenneth W

(2001) Leadership and Pandering: a theory of executive policymaking. American Journal of Political Science 45(3): 532–550.

Converse

Philip E

(2006) Researching Electoral Politics. American Political Science Review 100(4): 605–612.

Della Vigna

Stefano

Kaplan

Ethan

(2007) The Fox News Effect: media bias and voting. Quarterly Journal of Economics 122(August 2007): 1187–1234.

Durante

Ruben

Knight

Brian

(2012) Partisan Control, Media Bias, and Viewer Responses: evidence from Berlusconi’s Italy. Journal of the European Economic Association 10(3): 451–481.

10.

Enikolopov

Ruben

Petrova

Maria

Zhuravskaya

Ekaterina

(2011) Media and Political Persuasion: Evidence from Russia. American Economic Review 101(7): 3253–85.

11.

Feddersen

Timothy J

Pesendorfer

Wolfgang

(1996) The Swing Voter’s Curse. American Economic Review 86(3): 408–424.

12.

Furedi

Frank

(2005) Politics of Fear: Beyond Left and Right. London: Continuum Press.

13.

Gardner

Dan

(2009) Risk: The Science and Politics of Fear. London: Virgin Books.

14.

Gentzkow

Shapiro

Jesse

(2004) Media, Education, and Anti-Americanism in the Muslin World. Journal of Economic Perspectives 18(3): 117–133.

15.

Gerring

John

(2007) Is There a (Viable) Crucial-Case Method? Comparative Political Studies 40(3): 231–253.

16.

Hallin

Daniel C

Mancini

Paolo

(2004) Comparing Media Systems: Three Models of Media and Politics. Cambridge: Cambridge University Press.

17.

Holland

Paul W

Wainer

Howard

(1993) Differential item functioning. Hillsdale, NJ: Erlbaum.

18.

Howlett

Michael

Ramesh

Pearl

Anthony

(2009) Studying Public Policy: Policy Cycles and Policy Subsystems. Oxford: Oxford University Press.

19.

ITANES (2001) Perché ha vinto il centro-destra. Bologna: Il Mulino.

20.

Kingdon

John W

(1995) Agendas, Alternatives and Public Policies. Boston, MA; Toronto, Canada: Little, Brown and Company.

21.

La Ferrara

Eliana

Chong

Alberto

Duryea

Suzanne

(2008) Soap Operas and Fertility: Evidence from Brazil. Università Bocconi, mimeo. Available at: http://www.iadb.org/res/files/WP-633updated.pdf.

22.

Larcinese

Valentino

Puglisi

Riccardo

Snyder

James

(2011) Partisan Bias in Economic News: evidence on the agenda-setting behavior of U.S. newspapers. Journal of Public Economics 95(9/10): 1178–1189.

23.

Lawson

Chappel

McCann

James A

(2004) Television News, Mexico’s 2000 Elections, and Media Effects in Emerging Democracies. British Journal of Political Science 35(1): 1–30.

24.

Lazarsfeld

Paul

Berelson

Bernard

Gaudet

Hazel

(1944) The People’s Choice: How the Voter Makes up His Mind in a Presidential Campaign. New York, NY: Columbia University Press.

25.

Legnante

Guido

Baldassarri

Delia

(2010) Campagne Elettorali e Mediazione Sociale: esposizione ai media e relazioni interpersonali. In Bellucci

Paolo

Segatti

Paolo

(eds) Votare in Italia: 1968–2008. Bologna, Italy: il Mulino.

26.

L’Ulivo per Rutelli – insieme per l’Italia (2001) Rinnoviamo l’Italia, insieme. Il programma dell’Ulivo per il governo 2001–2006 presentato da Francesco Rutelli. mimeo. Available at: http://www.perlulivo.it/politiche2001/programma.pdf

27.

Lupia

Arthur

Menning

Jesse O

(2009) When can politicians scare citizens into supporting bad policies? American Journal of Political Science 53(1): 90–106.

28.

Mancini

Paolo

(ed) (2003) La Posta in Gioco. Temi, Personaggi e Satira nella Campagna Elettorale 2001. Rome, Italy: Carocci.

29.

Mancini

Paolo

(2009) Elogio della Lottizzazione: La Via Italiana al Pluralismo. Rome-Bari, Italy: Laterza.

30.

Marchese

Mirella

Milazzo

Giuseppe

(2001) L’Agenda dei Telegiornali sulle Notizie di Criminalità e Immigrazione: Un Confronto tra il 2000 e il 2001. Osservatorio di Pavia, mimeo. Available at: http://www.osservatorio.it/download/criminalita.pdf.

31.

Marletti

Carlo

(2002) Il ciclo dell’antipolitica e i risultati delle elezioni del 13 maggio in Italia.Verso un nuovo clima d’opinione?. Comunicazione Politica 3(1): 9–30.

32.

McCombs

Maxwell E

(2005) A Look at Agenda Setting: past, present and future. Journalism Studies 6(4): 543–557.

33.

McCombs

Maxwell E

Shaw

Donald L

(1972) The Agenda Setting Function of Mass Media. The Public Opinion Quarterly 36(2): 176–187.

34.

Ministero dell’Interno (2006) Rapporto sulla Criminalità. Rome, Italy: Ministero dell’Interno. Available at: http://www1.interno.gov.it/mininterno/export/sites/default/it/assets/files/14/0900_rapporto_criminalita.pdf.

35.

Moisi

Dominique

(2009) The Geopolitics of Emotion: How Cultures of Fear, Humiliation, and Hope are Reshaping the World. New York, NY: Anchor Books.

36.

Newell

James

(ed.) (2003) The Italian General Election of 2001: Berlusconi’s Victory. Manchester: Manchester University Press.

37.

Petrocik

John R

(1996) Issue Ownership in Presidential Elections, with a 1980 Case Study. American Journal of Political Science 40(3): 825–850.

38.

Puglisi

(2011) Being the New York Times: the political behaviour of a newspaper. B.E. Journal of Economic Analysis & Policy, 11(1), (Contributions), Article 20. Available at: http://eprints.lse.ac.uk/19292/1/Being_the_new_york_times.pdf.

39.

Robinson

James A

Torvik

Ragnar

(2008) The Real Swing Voter’s Curse. American Economic Review: Papers & Proceedings 99(2): 310–315.

40.

Wanta

Wayne

Ghanem

Salma

(2007) Effects of Agenda Setting. In Preiss

Gayle

Burrell

Allen

Bryant

(eds) Mass Media Effects Research: Advances Through Meta-Analysis. Mahwah, NJ: Lawrence Erlbaum Associates, 37–51.

41.

Weaver

David H

(2007) Thoughts on Agenda Setting, Framing and Priming. Journal of Communication 57(1): 142–147.

42.

White

Stephen

Oates

Sarah

McAllister

Ian

(2005) Media Effects and the Russian Elections, 1999–2000. British Journal of Political Science 34(2): 191–208.

43.

Witte

Kim

(1992) Putting the Fear Back into Fear Appeals: the extended parallel process model. Communication Monographs 59(4): 329–349.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.11 MB

0.34 MB