• Referendum campaigns;
  • Media fairness;
  • Press coverage;
  • Welfare state;
  • Switzerland


  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Theoretical background: Criteria of fair media coverage
  5. 3. Methods and measurements
  6. 4. An overall view of press performance
  7. 5. A dynamic view of independence, inclusiveness, bias, and substance
  8. 6. A particularistic view of media fairness
  9. 7. Spatial patterns of media coverage
  10. 8. Conclusion
  11. References
  12. Appendix

Abstract:  The mass media are assigned an important role in political campaigns on popular votes. This article asks how the press communicates political issues to citizens during referendum campaigns, and whether some minimal criteria for successful public deliberation are met. The press coverage of all 24 ballot votes on welfare state issues from 1995 to 2004 in Switzerland is examined, distinguishing seven criteria to judge how news coverage compares to idealized notions of the media’s role in the democratic process: coverage intensity, time for public deliberation, balance in media coverage, source independence and inclusiveness, substantive coverage, and spatial homogeneity. The results of our quantitative analysis suggest that the press does fulfil these normative requirements to a reasonable extent and that fears about biased or deceitful media treatment of ballot issues are not well-founded. However, some potential for optimizing the coverage of referendum campaigns by the Swiss press does exist.

1. Introduction1

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Theoretical background: Criteria of fair media coverage
  5. 3. Methods and measurements
  6. 4. An overall view of press performance
  7. 5. A dynamic view of independence, inclusiveness, bias, and substance
  8. 6. A particularistic view of media fairness
  9. 7. Spatial patterns of media coverage
  10. 8. Conclusion
  11. References
  12. Appendix

Instruments of direct democracy are at the heart of the Swiss political system. They are widely used, so much in fact that Switzerland alone accounts for half of all referendums held at the national level all over the world (Kaufmann et al. 2005; DuVivier 2006). Quite expectedly, this unequalled degree of direct democratic practice has triggered research to explore the determinants, conditions, and outcomes of the use of initiatives and referendums, as well as research to find out the causes of their success or failure. For example, Swiss scholars have directed their attention to the institutional effects of the referendum on the integration of political forces and to its role in shaping the Swiss concordance system (e.g., Neidhart 1970; Papadopoulos 1998; Vatter 2002). Another important area of inquiry focused on individual and aggregate citizen behaviour in referendum votes, relating voting patterns to the level of elite support or to structural properties of campaign propaganda (e.g., Hertig 1982; Trechsel and Sciarini 1998; Bützer and Marquis 2002).

In this study, we follow a complementary approach. Specifically, we ask how political issues are communicated to citizens during referendum campaigns, and whether some minimal criteria for successful public deliberation are met. Our analysis is based on the press coverage of twenty-four ballot votes on welfare issues, spanning more than two legislative periods (1995–2004) of highest importance for the development of the Swiss welfare state. Accordingly, our aim is to shed light on the journalistic perception of ballot campaigns. However, the views of political actors directly involved in partisan campaigns (e.g., parties, business associations) are also reflected in their agenda-building efforts to capture and manage media attention. It is from the interplay between these two perspectives on the news process — the journalists’ gatekeeping role and the elites’ agenda-building role — that the focal questions of this study arise. What is the degree of media autonomy? How inclusive are journalists’ accounts of the political actors involved in campaigns? How biased are their portrayals of issues?

In the next section of the paper we present some current ideas about public deliberation and provide theoretical insights into the structure and functions of referendum campaigns. We thus set a number of criteria for assessing whether media campaigns may be regarded as wise guides to sound decision-making. Section 3 then presents the empirical data collected and the methods used for measuring the quality of media campaign coverage. Sections 4 to 7 offer evidence of how journalists report on welfare state issues. We show that the press treatment of these issues does not fall short of expectations as concerns seven distinct criteria of “fair coverage”. On this overall basis, we conclude in Section 8 that some fears about skewed or deceitful media treatment of ballot issues are not justified, but also that the communication and framing of issues does not meet all conceivable normative requirements.

2. Theoretical background: Criteria of fair media coverage

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Theoretical background: Criteria of fair media coverage
  5. 3. Methods and measurements
  6. 4. An overall view of press performance
  7. 5. A dynamic view of independence, inclusiveness, bias, and substance
  8. 6. A particularistic view of media fairness
  9. 7. Spatial patterns of media coverage
  10. 8. Conclusion
  11. References
  12. Appendix

This article focuses on the “supply side” of referendum campaigns. Specifically, we attempt to define a number of dimensions on which to judge the fairness and democratic utility of the coverage of referendum issues in the Swiss media. The importance of articulating “democratic expectations of media performance” has been emphasized in several contributions (e.g., Gurevitch and Blumler 1990; McLeod et al. 2002). In particular, the “social responsibility” of the media has attracted much attention, leading scholars to define a number of desirable qualities of media content –– some of which may be properly regarded as “ethical”, but others are more directly related to the functioning of democratic systems (e.g., Bunton 1998; Christians and Nordenstreng 2004).

Our questioning in this article takes place in the interplay of several disciplines — including media ethics and the theory of deliberative democracy. These are broad and currently expanding disciplinary areas rather than formalized theories coupled with established methods and fieldwork, as the example of deliberative theory suggests (see Bächtiger et al. 2010). These theoretical accounts do not provide a clear and agreed-upon indication of what media fairness should be conceived of, let alone how to measure fairness. However, to the extent that the media are an important arena where democratic deliberation may occur, one major conceptual distinction from deliberative theory is helpful for framing our approach to media fairness.

Scholars in the deliberative framework usually advance that democratic decision-making and deliberation may be appraised (and purportedly justified) either from an epistemic or from a procedural standpoint (see Cohen 1986; Estlund 1997; List and Goodin 2001). The distinction is captured in the question of “whether we want our political outcomes to be right or whether we want them to be fair” (List and Goodin 2001: 277). To be sure, democratic processes oriented toward the search for “truth” or toward the observance of procedures may often bring about similar outcomes. It is nevertheless easy to imagine cases where “correctness” lacks procedural fairness (e.g., one individual “rightly” chooses for all others) and other cases where procedural rigidity is misplaced from procedure-independent standards (e.g., it violates some notion of the “common good”).

The field of media ethics is concerned with the same kind of questions, asking whether and how the “fairness” goal of media coverage can put up with the sometimes dissonant ideal of the “correctness” of the views being presented — beyond the mere question of information “accuracy”. In this article we adopt a procedural stance and examine how media coverage fulfils a number of formal rules regarding the form, provenance, and diffusion of information.2 Even though correctness criteria may appear to involve considerations of a more normative nature, it is however obvious that any set of formal-procedural rules is itself not “given”. As will become clear, some of the procedural criteria described below have been the subject of considerable speculation and debate. Moreover, “fairness”, as the pivotal concept of procedural approaches, is to be taken here in a generic sense, as it extends beyond the notions of fairness developed in careful theoretical analyses of “democratic proceduralism” (Cohen 1989; Estlund 1997).3 This enlarged concept of media fairness posits that, whatever the media induce citizens to think and choose, they must do so within certain procedural boundaries.

In this article we will focus on seven criteria that appear particularly relevant to evaluating the fairness of the media coverage of referendum campaigns in Switzerland. In this and the following sections we identify these criteria at a fairly abstract level, based on the available literature, and we show how each of them provides an appropriate dimension for assessing the achievements of campaign coverage from a normative perspective. Our goal in this contribution is primarily comparative, as we aim at showing how fairness varies across issue domains, across media outlets, over time, and across geographical units. Therefore, we deliberately avoid setting precise boundaries between “fair” and “unfair” coverage. On the other hand, the measurement of relevant criteria should be as precise and abstracted from context as possible, in order to allow for valid and meaningful comparisons. Besides, in order to avoid conceptual ambiguities, let us make clear that the term “fairness” is used here to denote the “positive side” on each dimension of media coverage. Thus we do not wish to imply, for instance, that a short, two-week campaign coverage is deliberately “unfair” or intended to fool voters, but only that a longer time period is more desirable according to some formal-procedural normative standard.

Based on this broad definition, we now delineate seven specific criteria for appraising the fairness of media campaign coverage.

1. Sufficient media coverage of issues. We first consider the information function of the mass media and the public knowledge of political issues that is expected to stem from media information. Learning effects are extremely important from a normative perspective, since they attest to the capacity of the people to understand politics and to exert influence on policies, as emphasized by classical democratic theory (e.g., Berelson 1952; Kelley 1960; Krouse and Marcus 1984; but see Pateman 1970). As a matter of fact, it has long been shown that media exposure generates knowledge about issues and candidates, and enhances opinion strength and attitude integration (e.g., Berelson et al. 1954: chap. 11; Delli Carpini and Keeter 1996: chap. 5; Jerit et al. 2006). These effects are usually not overwhelming, but they are quite robust across a wide range of contexts. Accordingly, learning effects have been observed in the Swiss direct democratic context as well (Kriesi 1994; Kriesi 2005: chap. 4; Marquis 2006). Information holding, in turn, matters for political judgments and voting decisions (Bartels 1996; Sturgis 2003). For example, a lack of knowledge causes citizens to be more conservative on some issues and more liberal on others (Althaus 1998; Gilens 2001), and it usually prompts reliance on heuristic cues that can lead to serious misfits between the people’s actual choices and their own values and interests (e.g., Kuklinski and Quirk 2000). Therefore, the overall intensity of media campaign coverage probably matters both for the public knowledge of issues and for the final outcome of the ballots. In fact, few would question that a sufficient amount of information is necessary for the proper functioning of democratic institutions. However, as argued below, the content of information also deserves attention, especially as concerns value-laden information (e.g., Hofstetter et al. 1999).

2. Sufficient time for public deliberation. Time is needed for the public to become informed about the issues and to ponder the pros and cons of a ballot proposal. If all campaign coverage takes place in a last-minute avalanche, then it is unlikely to benefit citizens, however extensive it may be in terms of sheer volume. Quite logically, the influence of a campaign usually concentrates on citizens who make up their minds during that campaign, that is, for people whose voting decision was neither clear from the outset nor delayed until the very last campaign days (e.g., Chaffee and Choe 1980; Fournier et al. 2004; Marquis 2006). Were campaigns so short-lived as to be virtually nonexistent for the purpose of deliberation, the collective decision-making process would be essentially reduced to the use of last-minute shortcuts (e.g., status quo bias) or of long-standing attitudes and prejudices (e.g., a personal bias against “big government”). The longer the referendum campaigns, the more likely it is that the same issues will be tackled and thus the same arguments will be repeated. In turn, this cumulative exposure may be expected to facilitate information acquisition by citizens (see Marquis and Bergman 2009). In other words, frequent exposure to campaign information enhances the accessibility of the relevant concepts (e.g., Higgins 1996; Förster and Liberman 2007) and heightens the likelihood that they will permeate the citizens' voting considerations.

3. No outright bias in media coverage. A relatively unbiased media coverage of issues may be stressed as a prerequisite for sound collective deliberation. This “impartiality” or “balance” assumption is more or less explicitly present in the writings of political philosophers such as Arendt, Elster, and Habermas, but also in concrete experiments designed to mimic ideal deliberative procedures (e.g., Yankelovich 1991: chap. 12; Luskin et al. 2002). Likewise, the norm of balance is deeply enmeshed in Western (primarily Anglo-American) culture and values (Schudson 2001). It looms large in the self-descriptions and perceived professional standards of journalists (e.g., Tuchman 1972; Patterson 1998) and in academic studies of media coverage unveiling their structural biases (e.g., Gitlin 1980; Hallin 1984; Entman 2004; Bennett et al. 2007).4 In addition, the development of “media watchdog journalism” monitoring media bias is, in part, reflective of the increasingly acute perception of the media bias problem in Western societies (Graber 2006; Hayes 2008). For our present purposes, such “issue neutrality” may come in several forms: the media may simultaneously present the arguments of both sides on an issue, or they may successively alternate between pros and cons (and thus act neutrally on an aggregate basis), or else they may restrict themselves to presenting “facts” and avoid matters of opinion. Meanwhile, the requirement that there be no outright bias in media coverage must not be taken to mean that the media should display no partisan bias, or that they should avoid expressing opinions altogether. Actually, some bias is inevitable and probably inconsequential, provided that it is not systematic and identical across all media outlets and issues. In addition, some mix of factual information and opinion statements is probably preferable to either one alone, for voters most certainly need –– and look for –– both types of information to make up their minds. As one study has shown, a voter’s general need for orientation toward media information is codetermined by her interest in acquiring knowledge about facts and about journalistic evaluations (Matthes 2005).

4. Source independence. Similar to balance, the “source independence” norm implies that the media are assigned responsibility for addressing issues in an autonomous manner. In this case, the normative expectation is that journalists should not depend too heavily on government, political parties and special interests for getting and framing information on campaign issues. Actually, there is a widespread belief in Western societies, including Switzerland, that journalists have “surrendered to complexity” and mainly serve as emissaries or “postmen”, as Wuerth (1999: 373) put it. According to the “determination hypothesis” and similar accounts of media-politics relationships, most news originates from official sources, including government and various branches of the administration, and especially from governmental public relations activities. Such source dependency, so the critics say, would lead to the formation of a misinformed and quiescent public (e.g., Baerns 1979; King and Schudson 1995; Shoemaker and Reese 1995; Bentele 2003). For instance, Grossenbacher et al. (2006) found that the press conferences held by two Swiss cantonal governments were indeed covered extensively by the local media, and that the press releases issued on these occasions were often published in original or only slightly abridged form, reproducing the local authorities’ positive self-assessments and issue priority. Likewise, more than half of the news stories in the U.S. press persistently emanate from official sources (e.g., Sigal 1973; Graber 2006). However, some scholars have cast doubt on the generalizability of such findings. For instance, it was found that the parties’ press releases rarely find their way into the German print media and/or that they are frequently altered and reframed by journalists (e.g., Donsbach and Wenzel 2002; Fröhlich and Rüdiger 2006). To some extent, journalists and their political sources are interdependent, both from a historical perspective (Schönhagen 2008) and in terms of role relationships (Merritt 1995). Journalists themselves are ambivalent about the autonomy issue. On the one hand, they increasingly value public relations sources and practitioners in the course of their professional experience (Sallot and Johnson 2006), not least because by creating long-term relationships with insiders they are granted routine access to valuable first-hand information. On the other hand, “autonomy” is seen as an important journalistic norm (e.g., Gurevitch and Blumler 1977; McDevitt 2003), admittedly with some important differences between national contexts (Patterson 1998; Statham 2006).5 In the present study, we will assess to what extent journalistic accounts of the campaign issues are elaborated “independently” or based instead on “self-serving” sources of information such as government, parties, and referendum committees.

5. Source inclusiveness. The norm of inclusion calls for the coverage of the whole diversity of viewpoints, arguments, and groups engaged in a referendum campaign. This norm thus concerns the question of how broad the variety of sources to which journalists refer (and which they make available to the public) should be. From another perspective, it asks how open or restricted the access to the media should be for different actors to voice their standpoints toward a referendum issue. Several democratic theorists have stressed the importance of an unrestricted and equal access for all individuals and societal groups to democratic processes in general (e.g., Dahl 1989: 119–131) and to the processes of public deliberation and will-formation in particular. The inclusiveness of the latter is especially prominent in theories of deliberative democracy. Cohen (1989: 21–23, 30; see also Dryzek 1990; Habermas 1992) regards inclusiveness and open access as a crucial aspect of an “ideal deliberative procedure”.6 However, various empirical studies have raised doubts about whether the media in reality live up to these expectations of democratic theorists (e.g., Page 1996; Gerhards 1997; Graber 2003). Pfetsch (2004) and Berkowitz (2009) argue that the degree of inclusiveness heavily varies depending on factors such as the cultural context, the journalists’ social characteristics, or the issue at stake. Specifically for Swiss referendums, Marquis and Bergman (2009) report a quite marked decrease in the variety of actors involved in advertisement campaigns in the 1990s. One reason for a low variety of sources might be the increasing commodification and commercialization of the media, which presses them to restrict themselves to sources compatible with the mainstream (Meier and Jarren 2002). Second, an increasing professionalization in the PR departments of the more established political actors might raise the hurdles for less prominent and financially disadvantaged groups to find their way into the media (Gurevitch and Blumler 1990). Third, while there do exist strategies for political “outsiders” to get included in a mediated political discourse, established political actors have also developed counter-strategies to hold outsiders and their arguments off the media attention (Kriesi 2003: 221–225). In the following analysis, we will assess the variety of reported viewpoints by the number of different groups of actors which journalists draw on as information sources.

6. Substantive coverage. For many observers, it is not enough that media information be balanced and independent of special interests — in addition, it has to address the substance of political issues. In fact, an essential part of campaign news coverage is framed in terms of non-substantive features of elections, such as their “game” or “horserace” aspects, and ignores their underlying issues (e.g., Strömbäck and Dimitrova 2006). A strong case has been made against the pervasiveness and deleterious influence of horserace information in a variety of election campaigns (e.g., Gollin 1980; Bartels 1988; Patterson 1994). We therefore only briefly discuss the relevance of the issue for the Swiss direct democratic environment. As pointed out by Rothmayr and Hardmeier (2002) and Longchamp (1998), the practice of direct democracy in Switzerland has inhibited rather than stimulated the development and use of opinion polls by both the media and governmental agencies. This is because frequent and “real” voting results are widely considered to provide a better account of public opinion than “unrealistic”, “out-of-context” opinion polls. However, Swiss journalists may find such polls attractive to report, in light of the news media’s preference for the “horserace” aspects of campaigns and for warfare and sports narratives. Moreover, they may find it convenient to draw on such ready-to-use information, without paying too much attention to its technical validity and practical relevance (see Hardmeier 1999). As a way to gauge the extent of horserace coverage, we will compare the amounts of “campaign-oriented” and of “issue-oriented” information provided by the media.

7. Spatial homogeneity. In a multicultural society such as Switzerland, the protection of minorities is usually a pressing issue. In addition, there is widespread concern that direct democratic decisions might heighten the risk of some “tyranny of the majority” (e.g., Donovan and Bowler 1998; Hajnal et al. 2002). In Switzerland, differences between linguistic regions, particularly as concerns voting behaviour in referendums, constitute one of the most salient political cleavages (e.g., Kriesi 1998; Linder et al. 2008). This cleavage is also one that is clearly identifiable and frequently reported on in the news media, probably because it is conflict-laden and thus has a high news value (Kriesi et al. 1996; Hardmeier 2000: 388). The Swiss media system is itself highly segmented along linguistic lines (Kriesi et al. 1996; Wuerth 1999; Blum 2003; Tresch 2008), and even the public service media find it increasingly difficult to promote national cohesion, as this task is incompatible with economic imperatives (Meier and Schanne 1994: 37–39). In fact, concerns about majority rule are not substantiated by empirical evidence in general terms, as referendum voting patterns are much more homogeneous than heterogeneous across cantonal and linguistic units (Diskin et al. 2007). In some cases, however, these concerns may be justified, as suggested by parliamentary votes or referenda results on welfare state issues, showing support for welfare policies to be significantly higher in the French- than in the German-speaking region (e.g., Leuthold et al. 2007). To be sure, it is perfectly legitimate for journalists from different places and cultural backgrounds to pursue their own agendas, for example as they index their treatment of job issues to real or perceived unemployment rates in their own area. Likewise, journalists may choose different specific information “formats” (e.g., heavily biased “opinion” articles or more balanced “points of view” articles) depending on their cultural environment and on how parochial the interests at stake are (e.g., Robinson 1995: 360–362). Accordingly, there is no reason to expect equal issue coverage all over the country. However, the influence of the media on voters can differ between the linguistic areas (e.g., Kriesi 1994; Marquis 2006: 623–628). Hence, heterogeneity of media content may reinforce existing cleavages and (re)produce undesirable tensions between linguistic communities. In the cases where the majority group prevails, it may fuel the argument of a “tyranny of the majority”. In sum, small between-region variations in media coverage are arguably preferable to great variations.

3. Methods and measurements

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Theoretical background: Criteria of fair media coverage
  5. 3. Methods and measurements
  6. 4. An overall view of press performance
  7. 5. A dynamic view of independence, inclusiveness, bias, and substance
  8. 6. A particularistic view of media fairness
  9. 7. Spatial patterns of media coverage
  10. 8. Conclusion
  11. References
  12. Appendix

This study bears on the 24 ballot votes dealing with welfare state issues which were held in Switzerland between 1995 and 2004 (see Table A1 in Appendix). Our empirical data consists of press articles collected in twenty-eight daily or weekly newspapers during an eight-week period preceding each ballot. The papers together account for a daily circulation figure of about two million copies and they are quite representative of the various Swiss regions.7 However, they were selected mainly for availability reasons, as they are consulted on a daily basis by the collaborators of the Année Politique Suisse at the University of Berne, Switzerland, and their content is classified into fine-grained thematic categories. It was thus easy to find out all campaign-related articles and to code the relevant information.

Two levels of analysis: articles and issue statements. Overall 4303 articles were found and coded. From these, a one-quarter sample of 1088 articles was randomly selected for additional analysis of the campaign issue statements.8 A maximum of three issue statements per article were coded, with the selection occurring on the basis of internal importance within each article. In total, 2859 issues were coded out of our sample, making for an average of 2.63 issues per article.

Intensity and length. The number of articles and their size (in centimetres squared, then converted in number of standard newspaper pages) were used to determine the intensity of campaign coverage.9 Based on the date on which an article was published, the number of days remaining before voting day was used to compute a measure of the “average campaign day” (mean of days remaining for separate ballots or newspapers).

Source independence and inclusiveness. Up to five sources per article were coded to determine who the “speakers” were, i.e., which categories of individuals or groups were primarily involved in determining the content of an article. Multiple sources were considered; for example, when a journalist interviews a political leader, two separate sources are coded. Similarly, up to two sources were coded for each issue statement within an article. We thus control for the fact that the original issue arguments delivered by a given source are often embedded in the discourse of a different “speaker” through positive or negative references. For example, it can be the case that all issue statements contained in an article where “journalists” are identified as the speaker category are in fact attributable to other actors to which journalists refer (e.g., partisan sources, government, and employees’ associations). Both speakers and issue statements’ sources were coded into fourteen mutually exclusive categories: (1) parties, politicians, and other party-related groups or individuals; (2) committees; (3) journalists and press agencies; (4) employers’ organizations; (5) employees’ organizations; (6) economic associations; (7) national government; (8) local government; (9) science- and education-related groups; (10) health-related groups; (11) disabled people’s groups; (12) women’s and families’ groups; (13) other non-profit organizations; (14) other actors. Together, categories (1) and (2) represent what we call “partisan actors”.10

The importance of journalists (category 3) will be compared with that of the other actor categories (both as “speakers” and as sources of issue statements) to assess the degree of the press’“source independence”. Note, however, that the total for any two groups of actors may exceed 100 percent since, by definition, several actors may be considered as the “sources” of an article or issue statement.

Further, we used the same source categories to build our measure of inclusiveness. It is simply the number of discrete categories from which at least one source of issue statements is drawn. In some of the forthcoming analyses, we will satisfy ourselves with this first measure (NBCAT). However, in order to control for the fact that some sources may be only marginally involved in agenda-building (and inclusiveness may thus be overestimated), we also assessed the degree of “equality” in the total references to sources of issue statements. This was achieved through the Index of Qualitative Variation (IQV):

  • image

where k is the total number of source categories, N is the total number of (weighted) issue statements, and fi is the frequency of all issue statements attributable to source i (N and f may be computed either as absolute numbers or as proportions). IQV scores can be integrated with NBCAT into a standardized measure of inclusiveness: INC=(NBCAT/14) × IQV. Somewhat surprisingly, NBCAT and IQV scores are totally uncorrelated at the level of ballot measures or of separate media outlets (r < .06). In fact, restriction of range in IQV values (M = 0.87, SD = 0.04, N = 24 ballot measures) explains why NBCAT has a much greater impact in determining INC scores.11 It also justifies using NBCAT alone, for simplification purposes, when IQV scores barely vary in cross-temporal and spatial comparisons.

Substantive coverage. The “format” of each article was coded following a three-fold distinction: (1) “opinion articles” (i.e., editorials, interviews, op-eds, free columns, or letters to the editor); (2) “factual articles” (i.e., mere reporting); and (3) “horserace information” (i.e., voting cues by parties and groups, opinion polls and other predictions, coverage of campaign events). The percentage of articles of the “horserace” type is used to assess the amount of “non-substantive” information. Among “substantive articles”, the percentage of “opinion” and “factual” articles is used to shed light on the “orientation” and “information” functions of press coverage, respectively.

Bias in coverage. Five broad categories aimed to assess the overall thrust of each article, based on the general slant of its arguments: (1) predominantly pro (i.e., suggesting a ‘yes’ vote on the ballot question); (2) predominantly con; (3) neutral; (4) mixed, i.e., controversial and including both sides; and (5) no argument. Similarly, the bias of each issue statement was coded on the basis of the same categories (excepting the “no argument” category, which is irrelevant). The categories were collapsed across ballot measures or other relevant units of analysis (newspapers, language areas, etc.) and combined to produce a summary indicator of coverage bias:

  • image

where pro is the number of positive items, con is the number of negative items, neu is the number of neutral items, and mix is the number of controversial items.12 Unbiased coverage is denoted by a B value of 0, while B values of –1 and +1 indicate the highest possible bias against and in favour of some proposal, respectively.

However, B scores are not well adapted for comparing bias in coverage across several ballot measures, since the meaning of pros and cons depends on how the ballot question was framed. One strategy is to rely on absolute values (i.e., |B| scores) to get an estimation of the overall extent of bias, regardless of direction. Another strategy is to devise a measure that expresses bias in terms of endorsement or opposition of welfare state policies. This requires that “pro” and “con” categories be inverted for ballots in the “retrenchment” category, i.e., measures that seek to reduce welfare benefits or to limit welfare expansion. Thus an original B score of +0.2 remains unchanged for expansion measures but its polarity is inverted and shifted to –0.2 for retrenchment measures. We call this new measure B*. Computing B* scores also implies that ballot measures can be clearly identified as “expansion” or “retrenchment” measures. In other words, only media contents that can be clearly assigned to “pro-welfare” and “anti-welfare” positions should be considered. This leads us to remove three “populist” proposals (ballots 724, 732, and 781; see Table A1) from the analysis whenever media bias is analyzed by means of B* scores.13 Except in this case, however, and unless indicated otherwise, all indicators used in this study are based on the total number of ballots (i.e., 24).

Spatial homogeneity. Indicators for each of the six above criteria of media fairness can be compared between the two main cultural areas in order to assess the homogeneity of media coverage. Simple differences are calculated between values for the majority (German-speaking) cultural area and those of minority areas. The only Italian-speaking paper in our database (i.e., Corriere del Ticino) is considered together with the French-speaking outlets.14

Validity checks. Each of the two independent coders read half the selected articles and measured all particular aspects listed above. Intercoder disagreement was examined from a sample of all coded articles and settled through discussion, resulting in more fine-grained and univocal coding procedures. The issue statements of all the 1088 selected articles were then coded again. Finally, a sample of 24 articles (one for each proposal) was drawn to check for coding consistency. Intercoder reliability was found to be satisfactory (Cohen’s kappa=0.704).15

Weighting procedures. Except for analyses bearing solely on the sheer number of articles, two weighting procedures were applied. First, in keeping with the literature on public attention and response (e.g., Neuman 1990; Price and Zaller 1993), the size of articles was logged and used as a weighting factor to account for the notion that marginal returns of an increase in news volume are generally diminishing. That is, for example, the difference between a five-inch-squared, barely visible, short article and a half-page article is probably more consequential than the difference between a full-page and a double-page article. Second, the “internal length” of issue statements was used to weigh their importance within our sample.16

4. An overall view of press performance

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Theoretical background: Criteria of fair media coverage
  5. 3. Methods and measurements
  6. 4. An overall view of press performance
  7. 5. A dynamic view of independence, inclusiveness, bias, and substance
  8. 6. A particularistic view of media fairness
  9. 7. Spatial patterns of media coverage
  10. 8. Conclusion
  11. References
  12. Appendix

Table 1 gives an overview of six of the seven criteria of media fairness presented in the theoretical and measurement sections — the “spatial homogeneity” dimension shall be analyzed separately below.

Table 1.   Indicators of fairness for media coverage of 24 ballot issues
Project (R=referendum, I=initiative, CP=counterproposal)NumberaSurfacebInclusi-venesscAverage campaign daydArticle biaseIssue biase% Jour-nalists (articles)f% Parties (articles)f% Jour-nalists (issues)f% Parties (issues)f% Repor-tingg% Opiniong% Horse-raceg
571 R / Reform of pension system (1995)30196.70.69422.70.0750.14484.318.528.743.748.433.717.9
572 I /“For extending the pension system” (1995)18564.30.66622.7−0.062−0.14281.621.739.438.849.236.814.0
602 R / Labour (week-end/night work) (1996)31795.10.85823.0−0.006−0.04281.013.022.839.647.538.613.9
622 R / Unemployment insurance (1997)21661.50.61819.5−0.074−0.10878.815.235.240.450.842.27.0
643 I / Retirement age (1998)25864.70.76224.5−0.050−0.10378.622.320.442.850.641.48.0
654 R / Labour (night work, maternity) (1998)8319.80.43418.60.089−0.07187.
684 R / Disability insurance (1999)8221.10.48823.3−0.176−0.21078.621.817.147.545.641.213.2
685 R / Maternity insurance (1999)31291.70.72926.70.0540.24981.521.831.244.653.635.411.0
721 I / Retirement age for women (2000)15648.90.44426.4−0.098−0.15185.519.631.047.048.640.510.9
722 I /“Flexible retirement age” (2000)15648.90.44426.4−0.098−0.13985.519.631.
724 I /“Reduced hospital costs” (2000)7122.50.38922.0−0.255−0.48993.720.144.923.046.742.211.2
732 I /“Cheaper drugs” (2001)17145.10.58725.5−0.071−0.31589.
752 I /“Secure pension system” (2001)10023.80.50520.9−0.046−0.01383.
762 I / Reduction of work time (2002)18660.00.73226.4−0.064−0.12292.14.622.929.661.933.24.9
781 I /“Gold to pensions” (2002)23864.50.62826.2−0.086−0.20889.922.217.257.736.632.830.6
782 CP /“Gold to pensions” (2002)27574.90.75725.90.0750.16889.821.025.547.742.031.726.3
792 R / Unemployment benefits (2002)14547.50.67426.90.0640.15696.
802 R / Financing of hospital treatments (2003)9221.80.52517.60.1560.38781.821.315.535.345.141.813.1
815 I /“Health must remain affordable” (2003)15249.90.62926.5−0.070−0.00888.717.326.066.748.037.614.4
816 I /“Equal rights for disabled” (2003)16345.40.75424.50.0670.40682.416.718.
819 I /“Apprenticeship places” (2003)10929.00.64522.9−0.0460.00885.114.333.841.244.126.629.3
831 R / Increase of pension age for women (2004)17048.90.68327.40.0810.02692.021.613.157.858.331.710.0
832 R / Financing pensions through VAT (2004)13140.30.47527.20.1140.25293.121.021.842.558.034.37.7
844 R / Maternity insurance (2004)23461.90.83424.60.0910.45693.712.626.144.357.522.220.3
Mean (N = 24)179.3 (Σ=4303)52.0 (Σ=1248.1)0.62324.10.086h0.182h86.417.325.843.550.635.214.2
Std. dev.73.522.40.1312.80.0490.1385.
Means for categories of project (standard deviations in parentheses)NumberaSurfacebInclusi-venesscAverage campaign daydArticle biashIssue biash% Jour-nalists (articles)f% Parties (articles)f% Jour-nalists (issues)f% Parties (issues)f% Repor-tingg% Opiniong% Horse-raceg
  1. Notes:a: Number of articles. b: Surface in standard newspaper pages (one page=1247 cm2). c: Number of actor categories × IQV (see text for calculation details). d: Mean number of days remaining before voting day. e: B scores comprised between -1 (all articles/issues against project) and +1 (all articles/issues in favour of project); see text for calculation details. f: Sum of categories “journalists” (including press agencies) and “parties” (including referendum/initiative committees) may exceed 100%, since several actors may be considered as the “sources” of an article. g: Percentage of articles in each category (see text for definition). h: Means computed from absolute values for each ballot, i.e. |B|. i: Not considered here: counterproposal to the “gold” initiative (ballot #782; see Appendix).

Pensions (N = 10)197.0 (63.5)57.6 (19.1)0.606 (0.120)25.0 (2.1)0.079 (0.021)0.135 (0.070)86.3 (4.5)20.8 (1.2)25.9 (7.4)48.2 (6.6)49.3 (6.2)36.1 (3.6)14.6 (7.6)
Health (N = 4)121.5 (41.2)34.8 (12.8)0.533 (0.091)22.9 (3.5)0.138 (0.076)0.300 (0.179)88.3 (4.2)17.7 (3.6)28.6 (10.5)40.4 (16.1)50.9 (7.5)37.3 (5.8)11.8 (2.2)
Maternity (N = 2)273.0 (39.0)76.8 (14.9)0.781 (0.052)25.6 (1.1)0.073 (0.019)0.353 (0.103)87.6 (6.1)17.2 (4.6)28.6 (2.5)44.5 (0.1)55.5 (1.9)28.8 (6.6)15.7 (4.7)
Disability (N = 2)122.5 (40.5)33.2 (12.2)0.621 (0.133)23.9 (0.6)0.121 (0.054)0.308 (0.098)80.5 (1.9)19.2 (2.5)17.6 (0.5)46.8 (0.7)51.4 (5.8)35.2 (6.0)13.4 (0.2)
Labour regulation (N = 4)173.8 (91.0)51.0 (29.5)0.667 (0.155)22.8 (2.8)0.051 (0.030)0.061 (0.042)86.4 (4.0)8.8 (4.9)23.6 (6.7)36.3 (4.6)47.2 (9.5)33.0 (4.2)19.7 (10.8)
Unemployment (N = 2)180.5 (35.5)54.5 (7.0)0.646 (0.028)23.2 (3.7)0.069 (0.005)0.132 (0.024)87.4 (8.6)14.2 (1.1)28.7 (6.6)36.4 (3.9)57.8 (7.0)37.1 (5.1)5.1 (1.9)
Retrenchment proposal (N = 6)205.2 (83.5)61.8 (26.9)0.669 (0.110)23.8 (2.7)0.079 (0.050)0.114 (0.064)85.1 (6.7)17.2 (3.7)23.2 (7.3)43.6 (7.8)52.6 (6.8)36.6 (4.3)10.9 (4.8)
Populist proposal (N = 3)160.0 (68.6)44.0 (17.1)0.535 (0.104)24.6 (1.8)0.137 (0.083)0.337 (0.116)90.9 (2.0)18.1 (4.4)30.1 (11.4)39.1 (14.3)49.0 (11.2)34.2 (6.0)16.7 (9.9)
Expansion proposal (N = 15)172.8 (67.3)49.7 (20.0)0.622 (0.134)24.1 (2.9)0.079 (0.029)0.178 (0.140)86.0 (4.5)17.1 (5.9)25.9 (6.8)44.4 (8.7)50.2 (6.7)34.8 (5.5)15.0 (7.8)
Referendum (N = 11)i189.4 (87.7)55.1 (27.9)0.638 (0.137)23.4 (3.4)0.089 (0.045)0.191 (0.132)86.2 (6.2)16.7 (5.6)22.6 (6.8)42.1 (6.7)51.4 (7.8)35.1 (5.5)13.5 (7.2)
Initiative (N = 12)i162.1 (51.1)47.3 (14.6)0.599 (0.122)24.6 (1.9)0.084 (0.054)0.175 (0.149)86.3 (4.3)17.5 (4.9)28.7 (8.0)44.5 (11.7)50.7 (7.2)35.5 (5.2)13.8 (7.7)

Intensity of coverage

There is little reason to expect journalists to devote equal attention to all subject matters; some issues are of greater concern to them (and to their readers) and are thus more likely to find their way onto the media agenda. In fact, there is a good deal of variation in the intensity of media coverage of the various ballot issues (SD is about 22 full newspaper pages for an overall mean of 52 pages). As shown in Table 1, the intensity of press coverage varies between the specific policy fields subsumed under the heading “welfare state issues”. Maternity seems by far the most important category of issues; in contrast, health and disability are of less importance. Further, it seems that retrenchment proposals lead to more intense campaigns than expansion or populist proposals. This is understandable, since cutting back benefits may lead to more mobilization by target groups than the promise of developing benefits can possibly achieve. As noted above, though, the two proposals to introduce a maternity insurance are a notable exception in this regard as they elicited substantial press coverage, probably due to the fact that the constitutional mandate to implement a maternity leave program dated back to the end of World War II and thus the issue was a recurring one.17 Likewise, referendums are somewhat more covered by the press than initiatives, probably because they bear more frequently on retrenchment programs, but the difference is modest.

In brief, the intensity of press coverage appears to vary between the different policy fields, arguably as a result of the journalists’ intrinsic interest in them and due to the varying mobilization of political and civil society actors. But, to a large extent, campaign intensity is probably also exogenously determined by the uncertainty of the result of the ballot. Both campaigners and voters are more likely to get involved in referendums when the margin of victory is perceived to be small (e.g., Downs 1957; Kirchgässner and Schulz 2005). Likewise, in our dataset the overall intensity of press coverage is substantially correlated with the closeness of the voting results.18 The closer the (expected) outcome, the more journalists report on issues.

Source inclusiveness

One indicator of the fairness of the coverage of a ballot issue is how many different political actors are involved in the deliberation process. Our standardized measure of inclusiveness points to a respectable diversity in the sources of issue statements in press coverage (overall mean=0.62). It is highest for maternity issues, and lowest for public health issues, as well as for issues comprised in the “populist” category. We thus assume that inclusiveness depends on the nature of issues (see Pfetsch 2004: 87–93), as further suggested by the fact that campaign intensity varies between policy fields, as we have seen. Indeed, intensity was shown to depend strongly on inclusiveness, as far as advertisement campaigns are concerned (Marquis and Bergman 2009). Similarly, in press coverage, the correlation between our indicators of intensity and inclusiveness is fairly high (Pearson’s r = 0.76). In other words, the more political actors enter the field (and journalists’ accounts presumably reflect this increasing diversity), the more likely the issue at hand is to receive general attention from the press — whether because inclusiveness determines intensity or because both coverage features are determined by some higher-level concept such as “issue importance”.

Length of coverage

Campaigns in Swiss direct democracy, whether through paid advertisements or in press reporting, usually unfold in two phases: there is first an accelerating expansion of coverage and then a sharp decline in the last days before the vote (e.g., Marquis 2006: 429–438; Tresch 2008: 147–149). Based on all articles published during all 24 campaigns, a third-order polynomial function modelling this typical development accounts for 67 percent of the variance in the total daily surface of articles.19 Hence, at the aggregate level, the two-phase pattern fits the data quite well: referendum campaigns on welfare state issues develop in a typical way, similar to what has been observed for other types of issues.20 In addition, all policy fields follow roughly the same pattern — the only notable difference is that campaigns on health matters (including maternity and disability) reach their peak earlier than other types of issues.

These findings may dispel concerns that most media information on ballot issues is released too fast and too late for citizens to use it efficiently in their decision making. But actually how long is the press coverage of ballot issues? Drawing on our indicator of the “average campaign day”, the “average information” is published some 24 days before the voting date (i.e, it corresponds to the “mean average day” for all 24 ballots). In the same perspective, the average median day is 21.6; this means that, in a typical campaign, half of the total (weighted) sum of all information has been released three weeks before voting day, and the other half is to be released in the remaining weeks. In addition, there appears to be little difference between the various ballot issues — even though ballot issues that draw more intense coverage (pensions, maternity) also tend to be covered for a slightly longer time. To be sure, actual exposure to the flow of information delivered during referendum campaigns may greatly vary from one citizen to another, depending for example on which media outlet (if any) they use for their information (see below). But our findings are hardly compatible with the argument that citizens are too time-pressed to make up their minds and are not given a chance to learn what the ballot issues are all about.

Source independence

The question of media independence and autonomy is important, both from the perspective of media practices and ethics and for the purposes of deliberative processes. For example, Tresch (2008: 142–149) shows that “agenda-building” by political actors through press conferences, party meetings, and other public relations activities, constitutes a substantial part of media content in referendum campaigns. Even though such coverage is certainly not without merit for citizens’ information and orientation, an overly reliance on external agenda-building efforts may lead to media instrumentalization beyond that which stems from patterns of media ownership (see Hallin and Mancini 2004).

If we first consider journalists and partisan actors (i.e., parties and committees) in their role of “speakers” at the level of articles, it comes as no surprise that journalists outweigh partisan actors (86 percent vs. 17 percent, on average). More remarkably, the differences between the various ballot proposals are quite limited. Journalists were identified as “speakers” in 80 to 88 percent of all articles in each of the six policy fields, while the corresponding interval is 14 to 21 percent for partisan actors — with the notable exception of labour market policies, where the parties’ share is only 9 percent.21 As for governmental speakers, they appear in less than 10 percent of articles in all policy fields. (Note that the percentages for the various categories add up to more than 100 percent, since any article can draw on several types of speakers.)

At the level of issue statements, recall that a total of 2859 issues were coded out of our sample of 1088 articles. Parties and committees account for about 44 percent of all coded issue statements, journalists and press agencies for 26 percent, governmental actors for 17 percent, employers’ and employees’ organizations for 14 percent, and other types of actors for 20 percent (similar to articles, percentages do not add to 100 percent because several categories of actors can be involved as sources of one issue).22 Overall, then, journalists appear much less important as a source of issue statements than as a source of articles (i.e., as “speakers”), which makes perfect sense. Conversely, even though partisan actors are rarely directly involved as speakers (e.g., through interviews or op-eds), they largely succeed in influencing campaign news by having their issue statements reported in the press (Bonfadelli and Blum 2000; Tresch 2008).

Bias in media coverage

The extent of bias in press coverage can also be investigated at two levels: at the level of whole articles and at the level of issue statements. The analysis draws in both cases on the classification of newspapers’ content into partial (pro or con), neutral, and controversial categories. The resulting B scores theoretically vary between –1 (all content is against ballot proposal) and +1 (all content is in favour of proposal). B scores are generally higher for issues than for articles, because, for one thing, there are far less “neutral” and “controversial” items (about 27 percent) in issue statements than at the level of whole articles (about 68 percent).

A striking feature of our results in Table 1 is the relative neutrality with which Swiss journalists report on the issues, at least taken collectively— the question of differences between media outlets will be addressed below. The mean absolute B value is 0.09 for articles and 0.18 for issue statements. In fact, a good deal of this “directional thrust” in media reporting is concentrated on ballot issues dealing with health matters (public health, maternity, disability). For example, the initiatives for “reduced hospital costs” (2000) and “equal rights for disabled people” (2003), as well as the referendum to oppose paid maternity leaves (2004), stirred up considerable criticism or enthusiasm from journalists, leading them to take a clear position on these subjects. On closer inspection, it appears that the general type of ballot proposals combines with their subject matter. As a category, the three populist proposals of the years 2000–2002 (including hospital costs) were treated in a far less even-handed manner than retrenchment or expansion proposals, especially as concerns bias in issue statements.23 We may summarize these findings by saying that journalists, most of the time, report about ballot issues in a quite balanced way.

Interestingly, bias in press coverage is substantially related to several other features examined here (correlations of 0.4 or higher). To begin with, article bias decreases as coverage intensity and inclusiveness increase. In other words, imbalance in coverage looms large when journalists have little to report and rely on fewer sources. At the level of issue statements, a crucial link seems to be with the importance of partisan actors as sources of statements: the more parties and committees are referred to as a basis for issue analysis, the lesser the bias.

Using B* values we can also determine how media coverage is basically oriented toward welfare state schemes (i.e., the extent to which campaign news systematically supports welfare state expansion/protection rather than welfare state limitation/ retrenchment). In so doing, we obtain a mean B* value of 0.01 for articles and 0.06 for issues — the press thus displays a very slight pro-welfare bias. This strongly suggests that there is no overall systematic pro-welfare or anti-welfare bias in press coverage. However, differences exist between the various policy fields. While neutral with respect to pensions, labour market and unemployment, Swiss media outlets as a whole seem committed to more generous social protection programmes in the area of public health (B*=0.19 for issue statements), including maternity and invalidity (B*=0.35 and 0.31, respectively).

It should also be noted that editorials and interviews are more critical of the welfare state than other format types. B* scores (issue statements) for editorials and interviews are about –0.2 and –0.1, respectively, as compared to +0.1 or higher for reports, letters to the editor, and campaign news. Journalists opposed to welfare policies, it would seem, rely heavily on writing editorials and selecting anti-welfare interviewees. But this does not suffice to countervail the bulk of pro-welfare coverage that stems from “factual reporting” or coverage of campaign events. Besides, the general thrust of media coverage is highly dependent on which types of actors get their messages across in the various newspapers, since more often than not journalists are balanced in their own issue statements (B* = –0.02). For instance, in the many cases where the parties’ issues find their way into the media, center-left parties are much more likely to voice pro-welfare issues than are right parties (B* = 0.42 vs. –0.26). Similarly, the bias brought about by employees’ organizations counterbalances that of employers’ organizations (B* = 0.35 vs. –0.29). Overall, referendum committees are relatively neutral toward the welfare system (B* = 0.04), but the organizations representing specific population groups (retired people, disabled people, youth, women, families) are overwhelmingly pro-welfare. Interestingly, the bias exhibited by governmental sources is quite different depending on the level of government. Institutions and civil servants at the national level (e.g., federal councillors, senior officials) are clearly less supportive of the welfare system (or, at least, of welfare state expansion) than are their counterparts at the cantonal or local level (B* = –0.15 vs. 0.48).24

Substantive coverage

We focus here on the three main formats of campaign coverage (i.e., “opinion articles”, “factual reporting”, and “horserace information”) to determine how “substantive” campaign coverage is. Concerns are often expressed that media news that unduly revolves around the horserace and “game” aspects of campaigns has a deleterious influence on public deliberation and citizens’ decision-making. In fact, horserace information was present in all campaigns but it is quite weak (14 percent of the overall amount of news, on average) compared to “opinion articles” (35 percent) and “factual articles” (51 percent). Horserace coverage varies between a minimum of 5 percent for unemployment ballot issues and a maximum of 20 percent for labour regulation ballot issues. In comparison, the share of opinion articles varies between 29 and 37 percent depending on policy domains, while the share of factual articles varies between 47 and 58 percent. Thus, unlike in other contexts, the horserace is clearly not the leading theme in Swiss referendum campaigns.25

5. A dynamic view of independence, inclusiveness, bias, and substance

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Theoretical background: Criteria of fair media coverage
  5. 3. Methods and measurements
  6. 4. An overall view of press performance
  7. 5. A dynamic view of independence, inclusiveness, bias, and substance
  8. 6. A particularistic view of media fairness
  9. 7. Spatial patterns of media coverage
  10. 8. Conclusion
  11. References
  12. Appendix

The picture thus far supports the notion of rather “fair” journalistic practices, as media coverage appears to be relatively balanced, autonomous and substantive. We now examine the hypothesis that journalists are however not immune to the pressure of the political environment and that they become increasingly dependent on external and/or biased sources of information over the course of campaigns. In other words, we ask whether the “fairness” of campaign coverage is affected by agenda-building efforts of political actors. We thus investigate the dynamics of campaign coverage, focusing on four aspects examined above: source independence and inclusiveness, bias in coverage and substantive coverage. We believe that the time dimension can add to our understanding of how much journalistic output is autonomous, inclusive, unbiased, and substantive — and possibly also why it deviates from such norms.

Beginning with source independence, one interesting result is that journalists are the only actor category whose importance as a source of issue statements increases throughout the whole campaign period, even though their share as “speakers” actually declines (see Figure 1). In contrast, parties and committees become a bit less marginal as speakers (gaining on average some 10 percent of the total share in the last seven weeks). But, most notably, partisan actors remain a major source of issue statements for press journalists during the whole campaign period; their share varies between 33 and 61 percent, but without clear upward or downward trend over time.26 Just as significant, however, is the fact that the sheer number of actor categories used as sources of issue statements grows almost linearly during the campaign period — except in the very last weeks — and actually almost doubles (from 3 to 6) in the first six weeks (see again Figure 1, right-hand axis). To use our terminology, this shows that press coverage becomes more inclusive as time goes by.27 For example, actors that may be regarded as “outsiders” (science and health professionals, associations defending the rights of women/families/disabled people, non-profit organizations, and other similar actors) rise from an average number of 0.7 (8th week before vote) to an average of 1.5 (4th week) and manage to keep their share of issue statements at just under 20 percent throughout the campaign.


Figure 1.  Source independence and inclusiveness over time.

Download figure to PowerPoint

Taken together, these results suggest that the dynamics of campaign coverage on welfare state issues is hardly compatible with a radical understanding of the “determination theory”. Our analysis reveals that journalists maintain their control over the newsgathering process and that no single source of campaign issue statements holds sway. As Figure 2 shows, however, this does not mean that journalists downplay the influence of political actors — or even try to do so in a collective sense. In fact, press coverage increasingly takes the form of “opinion articles” over time, while the share of “factual reporting” sharply declines. In the eight-week period, opinion articles become at least twice more frequent in relative terms (from 22 to 52 percent), while the share of factual articles considerably decreases (from 69 to 26 percent). With two exceptions, all campaigns exhibit this pattern, whereby opinions gain prominence at the expense of the presentation of facts as voting day draws nearer. This may be related to another finding stressed above, namely that “speakers”, i.e. those actors accountable for the content of an article, comprise less journalists and more partisan actors as time goes by. As for horserace news, its salience fluctuates within a narrow 9–16 percent range –– with the understandable exception of the last campaign week, given the close proximity of the vote and mobilization efforts by parties and other groups.28 In sum, there does not seem to be a strong focus or even focalization on the horserace during referendum campaigns in Switzerland. To a large extent, the overall evolution in the format of information is understandable, and some observers may find it comforting from a normative perspective. It can be argued that, as a collective actor, journalists first proceed to present the facts and main arguments (i.e., information function) and then provide citizens with particular opinions to help them take position on the issues (i.e., orientation function).


Figure 2.  Information format over time (substantive vs. non-substantive content).

Download figure to PowerPoint

Likewise, the “tone” of media coverage changes over time. To begin with, articles become more “partial”, i.e., they increasingly take sides for or against the various ballot proposals (see Figure 3). The share of partial articles amounts to slightly more than 20 percent in the first three weeks and approaches 50 percent by the last two weeks.29 In contrast, the share of partial issue statements— even though much higher than that of articles in overall level — slightly decreases over time. To a large extent this stems from the fact that issue statements get more “controversial” (rather than more “neutral”) as voting day draws closer, and this is probably reflective of the growing recognition by journalists of the complexity of ballot issues. That the press becomes more committed to its opinion-giving and orientation function in the later stages of campaigns does not entail, however, that it provides an increasingly biased picture of the ballot issues. As shown in Figure 3 (right-hand axis), it is striking how little deviation there is from perfect neutrality (i.e., B*=0) throughout the eight campaign weeks, based either on articles or on separate issue statements. In fact, 87 percent of the weekly bias scores for articles are comprised in a range of ± 20 percent around the neutral value.30 In sum, the fact that campaigns become more partisan over time does not necessarily imply that one side comes to prevail over the other. More often than not media campaign coverage remains remarkably balanced until the end. As we have shown above, ballot campaigns usually heat up in the final weeks, as agenda-building efforts by partisan groups intensify. But even then there is very little to suggest that the press gives in to external pressure and leaves the “undecided” citizens or “late deciders” exposed to the unchecked influence of partisan or special interests.


Figure 3.  Bias in media coverage over time.

Download figure to PowerPoint

6. A particularistic view of media fairness

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Theoretical background: Criteria of fair media coverage
  5. 3. Methods and measurements
  6. 4. An overall view of press performance
  7. 5. A dynamic view of independence, inclusiveness, bias, and substance
  8. 6. A particularistic view of media fairness
  9. 7. Spatial patterns of media coverage
  10. 8. Conclusion
  11. References
  12. Appendix

Based on a number of formal criteria, the press coverage of ballot issues on welfare state issues appears relatively “fair”. However, such evidence may be somehow illusory if it stems from the aggregation of contradictory patterns from different types of media outlets. For example, a balanced, unbiased portray of ballot issues at the aggregate level may conceal much greater variation between individual papers, some of which may be slanted toward welfare state expansion while others may be committed to welfare retrenchment. As most citizens read only one newspaper on a regular basis, this would mean that a majority of potential voters are exposed to a one-sided communication flow. In addition, if the papers on one side of the fence (e.g., the “pro-welfare camp”) have a larger overall readership than those on the other side, then the societal balance of information would no longer be guaranteed, despite indications to the contrary such as those provided in the preceding sections. To investigate this question, we ask how the coverage of welfare state issues compares across media outlets. Table 2 displays the whole range of indicators used thus far and shows how they vary between the 28 media outlets included in this analysis.

Table 2.   Indicators of fairness for all articles published in 28 daily and weekly newspapers
NewspaperNumberaSurfacebInclusi-venesscAverage campaign daydArticle biaseIssue biase% Jour-nalists (articles)f% Parties (articles)f% Jour-nalists (issues)f% Parties (issues)f% Repor-tingg% Opiniong% Horse-raceg
Neue Zürcher Zeitung (NZZ)528115.20.89127.3−0.127−0.27981.014.734.635.335.946.217.9
Basler Zeitung32197.60.87326.00.0440.11881.420.923.145.250.934.614.5
Le Temps (March 1998 –)29383.10.84527.2−0.018−0.03093.614.223.838.753.727.019.3
St. Galler Tagblatt28680.10.86524.50.0110.15385.316.726.050.049.930.819.4
Aargauer Zeitung26377.70.66426.2−0.0320.02383.
La Liberté24573.50.87428.0−0.0230.11198.88.529.244.764.520.215.2
Corriere del Ticino22258.40.86617.3−0.023−0.01481.225.723.843.463.630.95.6
24 Heures20168.00.78324.60.0040.00991.410.816.742.361.129.39.6
Neue Luzerner Zeitung17956.90.81828.7−0.027−0.12890.025.222.747.147.335.017.6
Nouvelliste et Feuille d’Avis du Valais17042.60.84018.0−0.0380.04075.420.034.742.
Berner Zeitung16657.40.78126.10.0380.16491.017.815.956.355.032.112.9
Le Quotidien Jurassien16535.20.75016.80.1450.44854.635.610.562.737.356.56.2
Schaffhauser Nachrichten12233.50.48414.6−0.094−0.10672.333.027.743.932.142.225.7
Tribune de Genève10437.70.74522.10.0120.20692.211.825.744.667.221.611.1
Solothurner Zeitung9026.80.40921.90.063−0.12482.924.116.853.248.446.55.1
Südostschweiz/Bündner Zeitung8923.00.68720.60.0460.12691.98.722.858.159.727.812.5
Journal de Genève (1995 – Feb. 1998)6018.50.75217.0−0.1080.00386.614.624.524.048.634.117.2
Berner Tagwacht (1995 – 1997)3818.90.22230.90.4250.43582.811.267.430.237.651.610.7
Schweizerische Handelszeitung218.60.27738.70.1170.05181.29.461.424.544.152.03.9
Le Nouveau Quotidien (1995 – Feb.1998)186.30.20616.3−0.026−0.200100.06.833.366.755.924.020.1
Die andere Zeitung (DAZ) (1995 – 1997)76.00.19432.00.1060.000100.034.277.647.365.834.20.0
Die Weltwoche53.70.19115.7-0.3470.089100.
Mean (N = 21)h153.7 (Σ=4303)44.6 (Σ=1248.1)0.74924.0−.001.05085.518.523.845.452.134.213.7
Standard deviation121.131.30.1385.50.0620.1469.
Means for categories of newspapers (standard deviations in parentheses)NumberaSurfacebInclusi-venesscAverage campaign daydArticle biaseIssue biase% Jour-nalists (articles)f% Parties (articles)f% Jour-nalists (issues)f% Parties (issues)f% Repor-tingg% Opiniong% Horse-raceg
  1. Notes:a: Number of articles. b: Surface in standard newspaper pages (one page=1247 cm2). c: Number of actor categories × IQV (see text for calculation details). d: Mean number of days remaining before voting day. e: B* scores comprised between –1 (all articles/issues against welfare state preservation/extension) and +1 (all articles/issues in favour of welfare state preservation/extension); populist ballots removed; see text for calculation details. f: Sum of categories “journalists” (including press agencies) and “parties” (including referendum/initiative committees) may exceed 100%, since several actors may be considered as the “sources” of an article. g: Percentage of articles in each category (see text for definition). h: Newspapers with fewer than 50 articles are excluded; all papers retained (N = 28) for number, surface, and average campaign day. i: National papers: Blick, Le Temps, NZZ, Tages-Anzeiger, DAZ. Supraregional papers: Journal de Genève, Neue Luzerner Zeitung, Le Nouveau Quotidien, St. Galler Tagblatt, Südostschweiz. Cantonal papers: Aargauer Zeitung, Basler Zeitung, Bund, Berner Zeitung, Corriere del Ticino, L’Express, La Liberté, Nouvelliste, Le Quotidien Jurassien, Schaffhauser Nachrichten, Solothurner Zeitung, Tribune de Genève, 24 Heures, Berner Tagwacht. Weekly papers: Schweizerische Handelszeitung, Sonntags-Zeitung, Wochenzeitung, Weltwoche.

Nationali236.6 (184.9)61.6 (41.6)0.641 (0.271)28.0 (2.1)0.018 (0.083)−0.007 (0.153)91.2 (7.1)19.7 (9.3)41.1 (19.2)36.7 (7.0)51.5 (9.7)33.5 (7.1)15.0 (7.5)
Supraregionali126.4 (95.7)37.0 (27.3)0.666 (0.238)21.4 (4.7)−0.021 (0.051)−0.009 (0.138)90.8 (5.2)14.4 (6.5)25.9 (3.9)49.2 (14.3)52.3 (4.7)30.4 (4.1)17.4 (2.7)
Cantonali172.3 (74.6)51.7 (22.8)0.702 (0.190)22.9 (4.6)0.036 (0.121)0.097 (0.166)83.5 (10.6)19.4 (8.5)24.5 (13.6)47.8 (8.2)52.4 (11.1)36.4 (11.5)11.2 (5.5)
Weeklyi19.0 (10.0)8.1 (4.0)0.203 (0.048)26.0 (8.4)0.125 (0.364)0.220 (0.257)87.7 (7.5)9.0 (6.8)38.6 (24.3)29.0 (37.4)48.9 (21.2)38.2 (23.6)12.9 (15.7)

We will focus here on the three aspects of independence, bias, and substance. An inspection of individual values and standard deviations shows that there is indeed some variation between newspapers on these three criteria, but that it is rather limited. According to our bias measure B* for articles, the economy-oriented newspapers NZZ and Journal de Genève may qualify as “anti-welfare”, as they were, on average, predominantly lopsided against welfare state preservation/extension. In contrast, the left-wing paper Quotidien Jurassien appears as “pro-welfare”.31 At the level of issue statements, there is greater variation in B* scores, but few papers stand outside a ± 20 percent range around the neutral value (anti-welfare: NZZ; pro-welfare: Quotidien Jurassien and Tribune de Genève). In addition, few single ballots gave way to skewed information even in any separate outlet. Of all instances with a sufficient number of cases, 11 percent were rather pro-welfare and 8 percent were rather anti-welfare.32 Weekly (mostly Sunday) papers represent an exception to this overall moderation in dealing with welfare issues, as they appear slanted toward pro-welfare positions (but self-evidently not the right-wing Weltwoche). In sum, the concern that our finding of balanced coverage might be an artifact of aggregation is not substantiated by our data.

Our findings concerning the criteria of source independence and substantive coverage tell a similar story. Significant variation between media outlets is scarce, even though some peculiarities are worth pointing out. For example, journalists outweigh partisan sources in the Tages-Anzeiger, while in other outlets (e.g., Quotidien Jurassien) these two categories are almost equally present — as speakers or sources of issue statements. In general, journalists play a larger role in national papers than in regional and cantonal papers, which may have to do with the resources media organizations must invest to have their journalists closely cover ballot campaigns. Finally, the “format” of articles shows little difference between newspapers. Some of them are more preoccupied with opinion formation than with factual reporting (NZZ, Nouvelliste, Quotidien Jurassien, Schaffhauser Nachrichten), while the opposite holds especially for a couple of French-speaking papers (La Liberté, Tribune de Genève, L’Express). With one single exception, horserace information account for less than 20 percent of campaign coverage in all papers. In any account, then, the overall variation in source independence and information format is limited. The existence of such variation cannot be taken to mean that the press coverage of ballot campaigns in Switzerland is a mere collection of disjoint “specific campaigns” with potentially divergent consequences for different segments of the electorate.

7. Spatial patterns of media coverage

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Theoretical background: Criteria of fair media coverage
  5. 3. Methods and measurements
  6. 4. An overall view of press performance
  7. 5. A dynamic view of independence, inclusiveness, bias, and substance
  8. 6. A particularistic view of media fairness
  9. 7. Spatial patterns of media coverage
  10. 8. Conclusion
  11. References
  12. Appendix

We now turn to the issue of spatial homogeneity. Provided that the Swiss media system is highly fragmented along linguistic lines, is this segmentation reflected in region-specific “styles” of media coverage? More specifically, to what extent are the particular dimensions of media coverage that we analyzed thus far similar in the two main Swiss regions? As becomes clear from Table 3, differences are in general rather limited, and sizable differences are only found for particular types of ballot issues.

Table 3.   Interregional differences with respect to eight indicators of media fairness and six policy fields
 Intensity aInclusiveness (# of source categories)Average campaign dayArticle bias (|B| scores)Issue bias (|B| scores)Source of articles = journalistsSource of issues = journalistsImportance of horserace information
  1. Notes: The figures show the difference between the two main linguistic areas (German-speaking vs. French- or Italian-speaking). A positive (negative) difference means that a given quantity is higher (lower) in the German-speaking area. In parentheses are the reference quantities, i.e., those measured in the German-speaking area.

  2. a: Intensity = inline image, with circi=circulation of newspaper i, surfij=surface of article j (in pages) in newspaper i, and voters=people eligible to vote (in thousands).

  3. **: p<.05; *: p<.10 (two-tailed tests checking for differences between paired-sample means).

Total (N = 24)4.2 (9.2)2.0 (9.6)1.6 (24.8)0.009 (0.091)0.013 (0.207)−0.7 (86.3)6.1** (28.1)3.9 (15.7)
Pensions (N = 10)4.7 (10.3)1.9 (9.7)1.2 (25.3)−0.010 (0.075)0.011 (0.145)−1.5 (85.9)5.1 (27.7)5.7 (16.6)
Health (N = 4)2.5 (6.4)1.3 (7.8)0.3 (22.8)0.037 (0.157)0.125 (0.420)2.2 (89.1)6.0 (33.0)5.6 (14.2)
Labour regulation (N = 4)3.2 (8.3)2.5 (10.8)3.0 (23.8)0.052 (0.080)−0.045 (0.071)1.3 (86.7)13.1* (28.5)4.6 (21.0)
Unemployment (N = 2)3.1 (8.8)2.0 (9.5)8.9 (27.0)−0.036 (0.053)0.002 (0.120)6.8 (89.2)10.2 (30.9)3.9 (7.3)
Maternity (N = 2)9.3 (15.6)0.5 (10.5)2.2 (26.4)−0.035 (0.065)−0.236 (0.297)−4.0 (86.5)0.5 (28.7)−4.5 (14.2)
Disability (N = 2)2.7 (5.5)4.0 (9.5)−4.1 (23.9)0.047 (0.131)0.178* (0.365)−11.3 (78.9)−1.6 (16.0)−1.8 (14.0)

Campaign intensity as reported in Table 3 is a composite measure derived from the surface of articles, from the circulation figures of the newspapers, and from the number of people eligible to vote.33 The measure thus points out differences in the overall potential exposure of individual citizens and shows that campaigns are clearly more intense in the German-speaking region. The difference is particularly noticeable and systematic with respect to pension and maternity issues. As concerns campaign length, the “average campaign day” indicator suggests that overall differences are slight. Campaigns may last a bit longer in the German-speaking press, but mainly for measures pertaining to unemployment, and in any case not for measures on disability. Further, as would be expected from these differences in intensity and length, media campaigns are also somewhat more “inclusive” in the German-speaking press, where, on average, two additional actor categories are used as sources of issue statements.

Turning to coverage bias, we again note that differences between the regions are marginal. However, as already pointed out above, bias is more conspicuous when measured at the level of issue statements rather than at the level of whole articles. Although the overall difference is negligible, this stems from the fact that the small differences that do exist cancel each other out across the various policy fields. Thus coverage in the German-speaking press was more skewed on health and disability measures, while it was less skewed on maternity issues (most notably, French- and Italian-speaking papers had been much more enthusiastically endorsing the 1999 maternity leave project than had their German-speaking counterparts).

Next, we compared the independence of journalists from external sources of information in the two regions. On the whole, the differences are modest. As speakers, German-speaking journalists were less prominent than journalists from other regions on maternity and disability issues, but they were more “autonomous” on unemployment. Differences were somewhat more pronounced as regards the source of issue statements. Journalists were more likely to be the source of reported issues in the German-speaking press than in the French- and Italian-speaking press, especially for unemployment and labour regulation (but not for maternity and disability issues). For example, French-speaking journalists sometimes draw on scientific or educational sources (8 percent) for addressing labour and unemployment issues, while their German-speaking colleagues never do so (at least in our sample). These regional patterns may have at least two different causes. On the one hand, turning to external sources may occur because journalists are less sure where they stand (or should stand) on the issues and/or because they believe that public opinion on the issues is not crystallized yet and need orientating information from political sources. On the other hand, the observed regional gaps may simply be reflective of more routinized relationships between journalists and political actors in some policy domains.

The “format” of information was then considered, with a focus on horserace information. Marginal regional gaps were observed, as the share of horserace items rarely exceeded 20 percent in any region, but the differences were rather systematic. In general, information in German-speaking papers is more often of the “horserace” type than in French- or Italian-speaking papers (an overall 4 percent gap), and this difference holds for 17 of 24 ballot proposals. The largest between-region differences in the share of horserace articles are found for pension and health issues, but there was actually more horserace coverage in the French/Italian-speaking area on maternity and disability issues.34

8. Conclusion

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Theoretical background: Criteria of fair media coverage
  5. 3. Methods and measurements
  6. 4. An overall view of press performance
  7. 5. A dynamic view of independence, inclusiveness, bias, and substance
  8. 6. A particularistic view of media fairness
  9. 7. Spatial patterns of media coverage
  10. 8. Conclusion
  11. References
  12. Appendix

The aim of this study was to assess the fairness of the press coverage of referendums on welfare state issues in Switzerland. We distinguished seven dimensions of media coverage in order to determine how it compares to idealized notions of the media’s role in the democratic process. In this concluding section we summarize our analysis of media coverage quality and, whenever possible and sensible, we relate our results to those of other studies.

First, as regards the intensity of press coverage in daily newspapers, we found about eight articles per proposal and newspaper, amounting to two or three full pages of coverage. With only modest variation across votes (though variation is more pronounced across media outlets), such coverage might be considered as a “fair” performance.35

Second, turning to the question of campaign duration, our results show that the press coverage peaked two to three weeks before the voting day, and that half of the total information had been released some 22 days before the poll. Although comparable studies of European referendums are lacking (see Novik 2009: 13) we interpret this average duration as “appropriate”, since press coverage was relatively gradual and did not surge in the last campaign days.

Third, we investigated the bias in media coverage. Overall, we find a minimal to nonexistent bias at the level of press articles (i.e., the general thrust of news stories) toward pro-welfare state orientation. When looking at particular issue statements, the bias appeared larger but still not overwhelming. In fact, bias was largely restricted to matters of public health in the larger sense (hospital costs, price of drugs, maternity, disability, etc.) and was virtually absent from campaigns on such themes as pensions, labour regulation, and unemployment. The variation in balance among different newspaper outlets is rather limited. Concerning articles in daily newspapers, only the Journal de Genève and the NZZ adopted a rather “anti-welfare” position while the Quotidien Jurassien was lopsided toward the “pro-welfare” camp. However, compared to news balance in other referendum campaigns, such as Pilon’s (2009) report on the Ontario Provincial Referendum and Hobolt’s (2009: 186–189) analysis of the first Irish referendum on the Nice Treaty, we conclude that the reporting on welfare state issues in the Swiss print media generally provides little ammunition to those who suspect or condemn a systematic bias in media coverage.

Fourth, the relative independence of journalists from external (particularly partisan) sources was examined. Similar to coverage bias, the dimension of media autonomy was investigated at two levels: articles and issue statements. At the level of articles, it was found that journalists are by far the most important “speakers”. Although they can delegate their agenda-setting role to other actors by soliciting their opinions through interviews, op-eds, or free columns, they rarely do so — at least not until the last three or four campaign weeks. Hence, our analysis suggests that the Swiss newspapers’ progressive loss of attachment to parties and other political groups (Meier and Schanne 1994) may have enhanced their “autonomy”— a concept not to be confused with that of journalistic “objectivity”, which is outside the scope of this study. However, when evaluated at the level of issue statements, the press coverage of referendum campaigns appears in a somewhat different light. Without there being anything “unfair” on a priori grounds, media coverage appears to largely reflect the issue agenda of the major political forces involved in campaigns (i.e., political parties and committees, governmental agencies, professional associations, etc.). Altogether, these non-media sources account for about three times the number of journalistic issue items.36 In our view, there is no basic contradiction between the findings from these distinct levels of analysis; taken together, they are reflective of both main functions traditionally assigned to news media — an information and an orientation function.

The dominance of journalists as sources of articles suggests that press journalists function as gatekeepers, selecting who may intervene directly in the news process, rather than as mere providers of a “free forum” to which any group or individual is granted equal and unlimited access. On the other hand, the press coverage of ballot issues is of course reflective — and possibly “reflexive” in the Luhmannian sense (Neidhardt 1994) — of what the main political actors have to say. Therefore, media agenda-setting is not exogenous, but largely driven by the agenda of various political elites. As our measure of source inclusiveness suggests, Swiss press journalists tend to rely on a broad array of different actor categories, and media coverage may thus be said to be “inclusive”. However, this substantial diversity in news coverage does not allow us to rule out the gatekeeping hypothesis: it is questionable whether all actors who wanted to participate in the campaign debates finally found their way in news reporting.

Next, our data provided insight into the substance of media coverage. We found that the often-criticized “horserace” nature of election coverage is not prevalent in the context of Swiss referendum campaigns. For one thing, horserace information accounts for only about 13 percent of all news items — a tiny proportion compared, for instance, with U.S. presidential campaigns (Strömbäck and Dimitrova 2006), the 1995 Quebec secession referendum (see Pilon 2009), or even with more readily comparable campaigns such as the 2000 Danish referendum on the adoption of the Euro (De Vreese and Semetko 2002). In addition, although the number of “factual information” items exceeds that of “opinion” items, both are important categories of press coverage. This again suggests that information and orientation are important functions performed by the media, and that in assuming these functions the media go well beyond predicting which side is likely to win.

Lastly, spatial homogeneity was defined as an additional norm of media coverage. This norm is rooted in the perception that public opinion on welfare state issues frequently diverges across the main linguistic areas. Accordingly, distinctive media coverage threatens to reinforce the pre-existing cleavages and to undermine national cohesion by promoting a “tyranny of the majority”. It was found that regional gaps in the press coverage of campaigns were minimal and largely restricted to three criteria: intensity of coverage, horserace reporting and source independence. In the German-speaking region, campaigns were somewhat more intense, more inclined toward horserace information, and journalists were more often the source of issues than was the case in the other linguistic areas. This may reflect either economic or cultural differences of media organizations between the regions. But again, the hallmark of this analysis is similarity rather than discrepancy. Even when looking at specific welfare policy areas (pensions, health, etc.), press coverage appears — with few exceptions — strikingly similar across both regions, in line with the more detailed analysis of Swiss European policy by Tresch (2008).

In conclusion, the coverage of referendum campaigns on welfare state issues by the Swiss press appears reasonably fair when examined from a broad perspective. However, this appraisal of “satisfactory” (though certainly not “optimal”) media campaign coverage is based on quantitative indicators that may fail to capture many intrinsic qualities of media coverage. Admittedly, the standards used to assess the fairness of media coverage were not excessively high, and the conclusions we draw from our analysis await confirmation from independent datasets and studies, in particular from more qualitative-oriented research “putting qualitative flesh on quantitative bones” (Tarrow 2004: 176). Accordingly, we call for further testing of the normative framework developed in the present study and for replications of our analysis in other settings. In addition, there is clearly a need of going beyond the mere “procedural” approach to fairness developed here. Although they can be estimated by quantitative indicators, procedure-independent criteria of media fairness may be best defined through qualitative methods.

  • 1

     This research was carried out at the University of Berne in the framework of a project on the “Political Consequences of Attitudes Toward the Welfare State” funded by the Swiss National Science Foundation, whose financial support is gratefully acknowledged (grant #100012–108274). The authors would like to thank Klaus Armingeon and Nathalie Giger for their support, as well as Hans Hirter, head of the Année Politique Suisse series, for giving us access to its invaluable archive of Swiss press releases. We also thank two anonymous reviewers for their extremely helpful comments on an earlier draft of our manuscript.

  • 2

     By choosing a procedural rather than epistemic perspective, we do not mean to imply any superiority or higher desirability of the former perspective in dealing with our subject.

  • 3

     Nevertheless, the fact that the elaboration and selection of procedural criteria is no less subjective than that of epistemic principles does not mean that the fundamental distinction between the two types of considerations is blurred. Generally speaking, our approach to media fairness is blind to information which may be of uttermost importance from an epistemic perspective — the specific opinions and origins of information purveyed in media coverage. We do not say that journalists are not (or should not be) driven by an orientation toward the common good, as when they downplay opinions that are disruptive and threatening for the democratic order — for example, British and German journalists have totally ignored extreme right parties in their coverage of election campaigns in the 1980s and 1990s (Bornschier 2010: 174-5). We simply say that this reference to the explicit content and implicit intent of media coverage is not relevant for our present purposes.

  • 4

     Biased coverage may occur in relatively subtle ways, such as when the photographs of candidates in newspapers are more or less favourable, depending on the fit between the candidate’s and the paper’s political leanings (Barrett and Barrington 2005). However, no consensus can be found in the literature as to whether media coverage is structurally biased, for instance conservative-leaning vs. liberal-leaning, or pro- vs. anti-establishment. Interestingly, the public perception that the media are biased toward liberal views may not be due to “real-world” media coverage bias, but to media self-coverage about biased media content (Watts et al. 1999).

  • 5

     Further, journalists may feel committed to autonomy as a result of their own audiences’ support for the norm of impartial coverage (Hargreaves and Thomas 2002) and as a result of their commitment being challenged by “popular commentators” such as bloggers (Singer 2007). In fact, public trust in the media may be undermined by excessive governmental control (Connolly and Hargreaves Heap 2007).

  • 6

     The inclusion of the whole diversity of opinions and societal groups is also a core request of the so-called “difference democrats”, who emphasize that it enhances not only the legitimacy of the opinion-building process, but also the stock of social knowledge and information available to the participants in the public discourse, thus facilitating well-informed and rational decisions (Young 1990: 119; 2000: 81–120; see also Connolly 1991).

  • 7

     The 1998 circulation figures are taken from the WEMF marketing research institute. The two related outlets “Blick” and “Sonntags-Blick” are considered here as one and the same paper. In contrast, the French-speaking tabloid “Le Matin” was not available in the archive of the Année Politique Suisse, although it is one of the major Swiss papers. However, a glimpse at the data for “Blick”, its counterpart in the German-speaking region, suggests that tabloids play a relatively minor role in the coverage of referendum campaigns. As a matter of fact, the 28 outlets are edited in 16 different Swiss cantons (out of a total of 26 cantons). Further, according to survey data, a majority of people in all but three cantons (SO, TG, ZH) are “heavy” or “medium users” of at least one newspaper among those considered here ( [accessed: 15.02.2010]).

  • 8

     The selection was performed by simply picking every fourth article in the database, reinitializing the count for each ballot issue. Checks were made to ensure that the obtained sample was not different from the whole dataset as concerns provenance, format, and article bias. An issue statement is defined here as an argument or set of arguments related to one specific matter addressed in the campaign about a ballot measure. It may pertain to a policy theme (e.g., environment, energy), a driving value or principle (e.g., solidarity), distinctive qualities of the proposed measure (e.g., flexibility), or the larger context of the ballots (e.g., rationale for a tactic vote). Issue statements have been classified into thirty-three categories (full list available upon request from the authors). For reasons of space limitation, however, we will not dwell on the question of which specific issues were emphasized in the various referendum campaigns.

  • 9

     For various reasons (e.g., missing part), 24 articles could not be coded as to their size and were assigned the median value of all calculated sizes (266 cm2).

  • 10

     Unlike in other policy areas such as foreign policy (see Marquis 2006), virtually all committees on welfare state issues are backed by parties and politicians.

  • 11

     At the level of ballots and papers, INC is strongly associated with NBCAT (r > 0.96) but not with IQV scores (r < 0.19). This raises the question whether the aggregation rule (unweighted multiplication of the two measures) is appropriate, since both NBcat and IQV are important on theoretical grounds. At the empirical level, however, it makes sense not to give more weight to the variable with lesser variation.

  • 12

     We take into account “neutral” and “mixed” items because failing to do so leads to serious distortions in the cases where coverage is mostly neutral but where only a few biased stories would result in meaninglessly high scores of coverage bias. Thus, for example, a number of positive items twice larger than that of negative items but equal to the number of non-valenced (i.e., neutral and controversial) items yields a B of “only” about 0.14. By the same token, B scores are expected to be more polarized when measured at the level of issue statements rather than at the level of articles, because in the former case the number of non-valenced items is considerably lower. Note also that B is very strongly correlated (r > 0.98) with alternative measures of media bias, such as that used by Zaller and Chiu (1996).

  • 13

     For our purposes, we define populist proposals as emanating from right-wing nationalist parties and groups, and being opposed by other parties. They consist of “radical”, often “simplistic” solutions to welfare problems, usually charging “big business”, high-profit sectors, or available public wealth. However, as these proposals do not align with the traditional ideological cleavages on welfare issues, opposition to them does not imply an endorsement of the welfare system. Accordingly, these proposals are hardly comparable with other proposals.

  • 14

     Journalists and political actors from the two minority cultural areas usually share the same general orientation toward welfare politics (though not in other important policy fields) and are probably closer to each other than they are to journalists and political actors from the German-speaking majority. More importantly, differences between majority and all minority areas are more commented upon and likely to be more consequential for national cohesion than differences between minorities.

  • 15

     This is actually a more than “fair” performance according to the standard interpretation by Landis and Koch (1977: 165). 67 issue statements were coded by the two coders simultaneously, i.e., an average of 2.8 issues per article. Three issues were coded by only one coder and were removed from the analysis, thus very slightly overestimating intercoder reliability by obfuscating disagreement over the sheer number of issue statements present in a press article. However, when the three issues are taken into account, the level of disagreement rises by a very small amount, and κ is still 0.67.

  • 16

     Five rough categories approximating the “internal length” of issue statements were recoded into five ratios reflecting their share in the whole article: 0.20, 0.35, 0.50, 0.65, and 0.80.

  • 17

     Prior to the 1999 and 2004 votes, no less than three proposals to create a maternity insurance had already been rejected at the polls (Dec. 8, 1974; Dec. 2, 1984; and Dec. 6, 1987).

  • 18

     The correlation coefficient is –.54 for the number of articles and –.48 for total surface. Closeness is measured as the absolute difference between the actual result and a perfectly balanced result: closeness = | 50 –result |.

  • 19

     The following function was fitted to the data: Surface = 0.0015 time3– 0.1464 time+ 3.7937 time + 4.3653, with surface measured as a three-day moving average and time measured as the number of days remaining until voting day. The proportion of explained variance rises to 80 percent when Sunday editions are removed. Because of their non-continuous time structure, such editions introduce “noise” into the data. However, only five outlets are weekly papers or are published on Sunday, and only 2 percent of articles were published on that weekday.

  • 20

     At the level of each single ballot, the characteristic sigmoid shape can be found for all but four campaigns. With the exception of ballots 572, 622, 781, and 782, all functions have positive values for time3 and time, and negative values for time2. Article size (rather than number) was used, and time was recoded in eight weekly periods to avoid high fluctuation in daily values. The average R2 for all 24 campaigns is 0.61 (SD=0.23).

  • 21

     In this domain, parties are secondary to other representative bodies with which they often have strong ties, in particular employers’ and employees’ associations (15 percent).

  • 22

     These shares expectedly vary across policy fields. Journalists are most involved in unemployment, public health, and maternity issues (all 29 percent) and least involved with respect to disability (18 percent). Parties and committees are much quoted sources of statements on pensions and disability (48 and 47 percent) and less so on unemployment and labour regulation (36 percent).

  • 23

     The 2000 and 2001 initiatives (“reduced hospital costs” and “cheaper drugs”) were fought by all parties, while the 2002 initiative (“gold reserves to retirement pensions”) was launched by the Swiss People’s Party and backed by other smaller populist right parties. Accordingly, the latter was supported by an important minority of 46% of the popular votes, compared to 18% and 31% for the two earlier initiatives.

  • 24

     This difference is consistent across all types of issues, excepting maternity, and it may be explained in two ways. On the one hand, the Swiss federal political system is largely based on the “principle of subsidiarity”, which posits that the federal state should assume only those tasks which cannot be performed effectively at a more local level. However, the federal state assumes the greatest responsibility for managing social benefits and financing pension and health insurances, while local governments may feel less under pressure. In addition, cantons and municipalities have to draw on their own resources to provide social assistance to needy people who fall through the cracks of the social safety net and are no longer eligible for national welfare benefits. Accordingly, local authorities are less likely to endorse cutbacks in social benefits and are more inclined to support their constituencies’ demands for social services. On the other hand, the principle of collegiality implies that federal councillors from left parties are compelled to advocate the view of the government as a collective body even when it contradicts the position of their party or their own opinion. For example, Federal Councillor Ruth Dreifuss, member of the Social Democratic Party and head of the Federal Department of Home Affairs (1994–2002), often had to oppose initiatives from left groups aiming at expanding or introducing new welfare programmes.

  • 25

     This is all the more significant as the definition of horserace information used in this study is rather broad. As a matter of fact, voting cues by parties and major organizations account for the bulk of “horserace information” as operationalized here — rather than opinion surveys or descriptions of campaign events.

  • 26

     The same holds, albeit to a smaller extent, for governmental actors, who account for over one quarter of all issue statements in the first four campaign weeks, but whose importance then declines to less than 15 percent. Yet, as compared to their tiny share of 5 percent of articles as “speakers”, governmental actors do succeed in getting their issues on the campaign agenda, especially at the earlier stage.

  • 27

     For the sake of simplicity, Figure 1 displays only the number of issue source categories, because IQV values hardly vary (min: 0.85, max: 0.90) and thus the variations over time of the composite index INC boil down to variations in the number of source categories.

  • 28

     We may also add that most newspapers publish a summary list of all parties’ and important organizations’ voting cues in the last week before voting day (which we also included in the horserace category).

  • 29

     One natural hypothesis is that journalists themselves become less impartial over time. There is strong evidence supporting this assumption, as the share of partial items among journalists’ articles rises from a low 16 percent seven weeks before voting day to a full 44 percent in the last campaign week.

  • 30

     The same result is obtained if one takes absolute B scores (and hence all 24 campaigns are considered). However, when issue statements are considered instead of articles, the proportion of biased weekly values (|B|>0.2) rises to 52 percent. To some extent this stems from our sampling procedure, because many weekly values are derived from a limited number of cases — while the whole data is used for articles’ bias. Accordingly, compared to statistics based on exhaustive data, there is a heightened probability that more extreme values of issue bias are produced by chance alone. Moreover, these frequent deviations from neutrality are in general temporary and unlikely to last for longer than one week. In fact, the percent of adjacent weeks that exhibit a similar biased value (e.g., B scores greater than 0.2 for both weeks) is only 19 percent.

  • 31

     In the following, we will not comment on results for newspapers with a small number of published articles (N < 50). Note that the indicators in this section are computed on the basis of all articles published in each newspaper, without distinction of the specific ballot measures.

  • 32

     292 campaigns in individual papers were analyzed and B* scores for articles were computed for each of them. Only cases where a newspaper published more than 4 articles on a given campaign were considered. When the same procedure is applied to B* scores for issue statements, the proportion of seemingly “biased campaigns” is much higher, but also less reliable (see above note 30).

  • 33

    Le Nouveau Quotidien, DAZ, and Berner Tagwacht are not taken into account in this analysis, due to unavailable circulation figures, but they represent only about 2 percent of articles and surface. More consequential is the fact that the French-speaking tabloid Le Matin, which has the largest circulation in the western region, is not included in our data. The absence of this paper probably explains part of the difference between the two regions.

  • 34

     It may be that, due to a more skeptical opinion climate toward the latter issues in their region, German-speaking journalists had more incentives to engage in opinion formation rather than stick to hard facts and descriptions of the situation. This is especially true for maternity issues, as welfare state support is 30 percent higher on average among French- and Italian-speaking voters than among German-speaking voters (our own calculations from the municipality-level data available in the “Political Atlas of Switzerland”, Swiss Federal Statistical Office).

  • 35

     This rather lenient judgment may be qualified by comparing our results with those of Hobolt (2009: 207−208), who found more intense press coverage for the 2005 French and the Dutch referendum campaigns on the European Constitution, even if we restrict the analysis to the country’s two major broadsheet newspapers (NZZ and Le Temps, in the Swiss case) as Hobolt did. However, the figures might not be fully comparable; the vote on the European Constitution might have been of greater salience than single welfare state proposals. Moreover, the lower frequency of popular votes in France and the Netherlands as compared to Switzerland may explain a more intense coverage of those (rare) referendums. We should therefore await further research on news reporting in the Swiss context in order to compare our findings.

  • 36

     As compared to Tresch’s (2008: 142–149) findings on media coverage of European politics, state actors play a slightly less prominent role in debates on welfare state issues.


  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Theoretical background: Criteria of fair media coverage
  5. 3. Methods and measurements
  6. 4. An overall view of press performance
  7. 5. A dynamic view of independence, inclusiveness, bias, and substance
  8. 6. A particularistic view of media fairness
  9. 7. Spatial patterns of media coverage
  10. 8. Conclusion
  11. References
  12. Appendix
  • Althaus, S. (1998). Information Effects in Collective Preferences. American Political Science Review 92(3): 545558.
  • Bächtiger, A., S. Niemeyer, M. Neblo, M. Steenbergen and J. Steiner (2010). Disentangling Diversity in Deliberative Democracy: Competing Theories, Their Blind Spots and Complementarities. Journal of Political Philosophy 18(1): 3263.
  • Baerns, B. (1979). Öffentlichkeitsarbeit als Determinante journalistischer Informationsleistungen. Publizistik 24(3): 301316.
  • Barrett, A. and L. Barrington (2005). Bias in Newspaper Photograph Selection. Political Research Quarterly 58(4): 609618.
  • Bartels, L. (1988). Presidential Primaries and the Dynamics of Public Choice. Princeton: Princeton University Press.
  • Bartels, L. (1996). Uninformed Votes: Information Effects in Presidential Elections. American Journal of Political Science 40(1): 194230.
  • Bennett, L., R. Lawrence and S. Livingston (2007). When the Press Fails. Political Power and the News Media from Iraq to Katrina. Chicago: University of Chicago Press.
  • Bentele, G. (2003). Kommunikatorforschung: Public Relations. In Bentele, G., H.-B.Brosius and O.Jarren (eds.), Öffentliche Kommunikation. Handbuch Kommunikations- und Medienwissenschaft. Wiesbaden: Westdeutscher Verlag (5478).
  • Berelson, B. (1952). Democratic Theory and Public Opinion. Public Opinion Quarterly 16(3): 313330.
  • Berelson, B., P. Lazarsfeld and W. McPhee (1954). Voting. A Study of Opinion Formation in a Presidential Campaign. Chicago: The University of Chicago Press.
  • Berkowitz, D.A. (2009). Reporters and Their Sources. In Wahl-Jorgensen, K. and T. Hanitzsch (eds.), The Handbook of Journalism Studies. New York: Routledge (102115).
  • Blum, R. (2003). Medienstrukturen der Schweiz. In Bentele, G., H.-B.Brosius and O.Jarren (eds.), Öffentliche Kommunikation. Handbuch Kommunikations- und Medienwissenschaft. Wiesbaden: Westdeutscher Verlag (366381).
  • Bonfadelli, H. and R. Blum (2000). Helvetisches Stiefkind: die Rolle der Massenmedien bei der Vermittlung schweizerischer Aussenpolitik. Bern, NFP 42 Synthesis n° 28.
  • Bornschier, S. (2010). Cleavage Politics and the Populist Right: The New Cultural Conflict in Western Europe. Philadelphia: Temple University Press.
  • Bützer, M. and L. Marquis (2002). Public Opinion Formation in Swiss Federal Referendums. In Farrell, D. and R.Schmitt-Beck (eds.), Do Political Campaigns Matter? Campaign Effects in Elections and Referendums. London and New York: Routledge (163182).
  • Bunton, K. (1998). Social Responsibility in Covering Community: A Narrative Case Analysis. Journal of Mass Media Ethics 13(4): 232246.
  • Chaffee, S. and S.Y. Choe (1980). Time of Decision and Media Use During the Ford-Carter Campaign. Public Opinion Quarterly 44(1): 5369.
  • Christians, C. and K. Nordenstreng (2004). Social Responsibility Worldwide. Journal of Mass Media Ethics 19(1): 328.
  • Cohen, J. (1986). An Epistemic Conception of Democracy. Ethics 97(1): 2638.
  • Cohen, J. (1989). Deliberation and Democratic Legitimacy. In Hamlin, A. and P.Pettit (eds.), The Good Polity. Oxford: Blackwell (1734).
  • Connolly, S. and S. Hargreaves Heap (2007). Cross Country Differences in Trust in Television and the Governance of Public Broadcasters. Kyklos 60(1): 314.
  • Connolly, W.E. 1991. Identity, Difference: Democratic Negotiations of Political Paradox. Ithaca: Cornell University Press.
  • Dahl, R.A. 1989. Democracy and Its Critics. New Haven [etc.]: Yale University Press.
  • De Vreese, C. and H. Semetko (2002). Public Perception of Polls and Support for Restrictions on the Publication of Polls: Denmark’s 2000 Euro Referendum. International Journal of Public Opinion Research 14(4): 367390.
  • Delli Carpini, M. and S. Keeter 1996. What Americans Know about Politics and Why It Matters. New Haven and London: Yale University Press.
  • Diskin, A., A. Eschet-Schwarz and D. Felsenthal (2007). Homogeneity, Heterogeneity and Direct Democracy: The Case of Swiss Referenda. Canadian Journal of Political Science 40(2): 317342.
  • Donovan, T. and S. Bowler (1998). An Overview of Direct Democracy in the American States. In Bowler, S., T.Donovan and C.Tolbert (eds.), Citizens as Legislators. Direct Democracy in the United States. Columbus: Ohio State University Press (121).
  • Donsbach, W. and A. Wenzel (2002). Aktivität und Passivität von Journalisten gegenüber parlamentarischer Pressearbeit. Publizistik 47(4): 373387.
  • Downs, A. (1957). An Economic Theory of Democracy. New York: Harper Collins.
  • Dryzek, J.S. (1990). Discursive Democracy: Politics, Policy, and Political Science. New York: Cambridge University Press.
  • DuVivier, K. (2006). The United States as a Democratic Ideal? International Lessons in Referendum Democracy. Temple Law Review 79: 821876.
  • Entman, R. (2004). Projections of Power. Framing News, Public Opinion, and U.S. Foreign Policy. Chicago: University of Chicago Press.
  • Estlund, D. (1997). Beyond Fairness and Deliberation: The Epistemic Dimension of Democratic Authority. In Bohman, J. and W.Rehg (eds.), Deliberative Democracy: Essays on Reason and Politics. Cambridge: MIT Press (173204).
  • Förster, J. and N. Liberman (2007). Knowledge Activation. In Kruglanski, A. and E.T.Higgins (eds.), Social Psychology. Handbook of Basic Principles. New York: Guilford Press (201231).
  • Fournier, P., R. Nadeau, A. Blais, E. Gidengil and N. Nevitte (2004). Time-of-Voting Decision and Susceptibility to Campaign Effects. Electoral Studies 23(4): 661681.
  • Fröhlich, R. and B. Rüdiger (2006). Framing Political Public Relations: Measuring Success of Political Communication Strategies in Germany. Public Relations Review 32: 1825.
  • Gerhards, J. (1997). Diskursive versus liberale Öffentlichkeit. Eine empirische Auseinandersetzung mit Jürgen Habermas. Kölner Zeitschrift für Soziologie und Sozialpsychologie 49(1): 134.
  • Gilens, M. (2001). Political Ignorance and Collective Policy Preferences. American Political Science Review 95(2): 379396.
  • Gitlin, T. (1980). The Whole World is Watching. Mass Media in the Making and Unmaking of the New Left. Berkeley: University of California Press.
  • Gollin, A. (1980). Exploring the Liaison between Polling and the Press. Public Opinion Quarterly 44(4): 445461.
  • Graber, D. (2003). The Media and Democracy: Beyond Myths and Stereotypes. Annual Review of Political Science 6: 139160.
  • Graber, D. (2006). Mass Media and American Politics. Washington: CQ Press.
  • Grossenbacher, R., T. Forsberg, M. Koch and M. Brändli (2006). Politische Öffentlichkeitsarbeit in regionalen Medien. Kilchberg: Publicom AG.
  • Gurevitch, M. and J. Blumler (1977). Linkages between the Mass Media and Politics: A Model for the Analysis of Political Communications Systems. In Curran, J., M.Gurevitch and J.Woollacott (eds.), Mass Communication and Society. London: Edward Arnold (270290).
  • Gurevitch, M. and J. Blumler (1990). Political Communication Systems and Democratic Values. In Lichtenberg, J. (ed.), Democracy and the Mass Media. New York: Cambridge University Press (269289).
  • Habermas, J. (1992). Faktizität und Geltung. Beiträge zur Diskurstheorie des Rechts und des demokratischen Rechtsstaats. Frankfurt am Main: Suhrkamp.
  • Hajnal, Z., E. Gerber and H. Louch (2002). Minorities and Direct Legislation: Evidence from California Ballot Proposition Elections. Journal of Politics 64(1): 154177.
  • Hallin, D. (1984). The Media, the War in Vietnam, and Political Support: A Critique of the Thesis of an Oppositional Media. Journal of Politics 46(1): 224.
  • Hallin, D. and P. Mancini (2004). Comparing Media Systems. Three Models of Media and Politics. New York: Cambridge University Press.
  • Hardmeier, S. (1999). Political Poll Reporting in Swiss Print Media: Analysis and Suggestions for Quality Improvement. International Journal of Public Opinion Research 11(3): 25774.
  • Hardmeier, S. (2000). Meinungsumfragen im Journalismus: Nachrichtenwert, Präzision und Publikum. Medien und Kommunikationswissenschaft 48(3): 371395.
  • Hargreaves, I. and J. Thomas (2002). New News, Old News. An ITC and BSC Research Publication. London: ITC.
  • Hayes, A. (2008). Press Critics Are the Fifth Estate: Media Watchdogs in America. New York: Praeger.
  • Hertig, H.P. (1982). Sind Abstimmungserfolge käuflich? — Elemente der Meinungsbildung bei Eidgenössischen Abstimmungen. Annuaire Suisse de Science Politique 22: 3557.
  • Higgins, E. (1996). Knowledge Activation: Accessibility, Applicability, and Salience. In Higgins, E. and A.Kruglanski (eds.), Social Psychology. Handbook of Basic Principles. New York: The Guilford Press (133168).
  • Hofstetter, C. et al. (1999). Information, Misinformation, and Political Talk Radio. Political Research Quarterly 52(2): 353369.
  • Hobolt Binzer, S. (2009). Europe in Question: Referendums on European Integration. Oxford: Oxford University Press.
  • Jerit, J., J. Barabas and T. Bolsen (2006). Citizens, Knowledge, and the Information Environment. American Journal of Political Science 50(2): 266282.
  • Kaufmann, B., G. Kreis and A. Gross (2005). Direkte Demokratie und europäische Integration. Die Handlungsspielräume der Schweiz. Basel: Europainstitut.
  • Kelley, S. (1960). Political Campaigning. Problems in Creating an Informed Electorate. Washington: The Brookings Institution.
  • King, E. and M. Schudson (1995). The Press and the Illusion of Public Opinion: The Strange Case of Ronald Reagan’s ‘Popularity’. In Glasser, T. and C.Salmon (eds.), Public Opinion and the Communication of Consent. New York: The Guilford Press (132155).
  • Kirchgässner, G. and T. Schulz (2005). Was treibt die Stimmbürger an die Urne? Eine empirische Untersuchung der Abstimmungsbeteiligung in der Schweiz, 1981–1999. Swiss Political Science Review 11(1): 156.
  • Kriesi, H. (1994). Le défi à la démocratie directe posé par les transformations de l’espace public. In Papadopoulos, Y. (ed.), Présent et avenir de la démocratie directe. Geneva: Georg (3172).
  • Kriesi, H. (1998). The Transformation of Cleavage Politics: The 1997 Stein Rokkan Lecture. European Journal of Political Research 33(2): 165185.
  • Kriesi, H. (2003). Strategische politische Kommunikation: Bedingungen und Chancen der Mobilisierung öffentlicher Meinung im internationalen Vergleich. In Esser, F. and B.Pfetsch (eds.), Politische Kommunikation im internationalen Vergleich. Grundlagen, Anwendungen, Perspektiven. Wiesbaden: Westdeutscher Verlag (208239).
  • Kriesi, H. (2005). Direct Democratic Choice. The Swiss Experience. Lanham: Lexington Books.
  • Kriesi, H., B. Wernli, P. Sciarini and M. Gianni (1996). Le clivage linguistique. Problèmes de compréhension entre les communautés linguistiques en Suisse. Berne: Office fédéral de la statistique.
  • Krouse, R. and G. Marcus (1984). Electoral Studies and Democratic Theory Reconsidered. Political Behavior 6(1): 2339.
  • Kuklinski, J. and P. Quirk (2000). Reconsidering the Rational Public: Cognition, Heuristics, and Mass Opinion. In Lupia, A., M.McCubbins and S.Popkin (eds.), Elements of Reason. Cambridge: Cambridge University Press (153182).
  • Landis, J.R. and G. Koch (1977). The Measurement of Observer Agreement for Categorical Data. Biometrics 33(1): 159174.
  • Leuthold, H., M. Hermann and S. Fabrikant (2007). Making the Political Landscape Visible: Mapping and Analyzing Voting Patterns in an Ideological Space. Environment and Planning B: Planning and Design 34: 785807.
  • Linder, W., R. Zürcher and C. Bolliger (2008). Gespaltene Schweiz – geeinte Schweiz. Gesellschaftliche Spaltungen und Konkordanz bei den Volksabstimmungen seit 1874. Baden: Hier und Jetzt.
  • List, C. and R.E. Goodin (2001). Epistemic Democracy: Generalizing the Condorcet Jury Theorem. Journal of Political Philosophy 9(3): 277306.
  • Longchamp, C. (1998). Demoskopie: Seismograph oder Kompass? Ein Überblick über die Ausbreitung und Verwendung der politischen Umfrageforschung in der Schweiz. Referat vor der Erdöl-Vereinigung, 23.03.1998. Online: [accessed: 29.04.2010].
  • Luskin, R., J. Fishkin and R. Jowell (2002). Considered Opinions: Deliberative Polling in Britain. British Journal of Political Science 32(3): 455487.
  • Marquis, L. (2006). La formation de l’opinion publique en démocratie directe. Les référendums sur la politique extérieure suisse (1981–1995). Zurich: Seismo.
  • Marquis, L. and M. Bergman (2009). Development and Consequences of Referendum Campaigns in Switzerland, 1981–1999. Swiss Political Science Review 15(1): 6397.
  • Matthes, J. (2005). The Need for Orientation Towards News Media: Revising and Validating a Classic Concept. International Journal of Public Opinion Research 18(4): 422444.
  • McDevitt, M. (2003). In Defense of Autonomy: A Critique of the Public Journalism Critique. Journal of Communication 53(1): 155160.
  • McLeod, D., G. Kosicki and J. McLeod (2002). Resurveying the Boundaries of Political Communications Effects. In Bryant, J. and D.Zillmann (eds.), Media Effects. Advances in Theory and Research. Mahwah: Lawrence Erlbaum (215267).
  • Meier, W. and M. Schanne (1994). Medien-“Landschaft” Schweiz. Zurich: Pro Helvetia.
  • Meier, W. and O. Jarren (2002). Ökonomisierung und Kommerzialisierung von Medien und Mediensystem. Bemerkungen zu einer (notwendigen) Debatte. In Haas, H. and O.Jarren (eds.), Mediensysteme im Wandel. Struktur, Organisation und Funktion der Massenmedien. Wien: Braumüller (201216).
  • Merritt, D. (1995). Public Journalism and Public Life. Why Telling the News Is Not Enough. Mahwah: Lawrence Erlbaum.
  • Neidhardt, F. (1994). Öffentlichkeit, öffentliche Meinung, soziale Bewegungen. In Neidhardt, F. (ed.), Öffentlichkeit, öffentliche Meinung, soziale Bewegungen. Kölner Zeitschrift für Soziologie und Sozialpsychologie, Sonderheft 34 (741).
  • Neidhart, L. (1970). Plebiszit und pluralitäre Demokratie. Eine Analyse der Funktion des schweizerischen Gesetzesreferendums. Berne: Francke.
  • Neuman, W.R. (1990). The Threshold of Public Attention. Public Opinion Quarterly 54(2): 159176.
  • Novik, N. (2009). Do Campaigns Matter? The Contribution of European Union Campaigns to Broad Political Knowledge. Dissertation. Dublin: Trinity College.
  • Page, B.I. (1996). Who Deliberates? Mass Media in Modern Democracy. Chicago: University of Chicago Press.
  • Papadopoulos, Y. (1998). Démocratie directe. Paris: Economica.
  • Pateman, C. (1970). Participation and Democratic Theory. Cambridge: Cambridge University Press.
  • Patterson, T. (1994). Out of Order. New York: Vintage.
  • Patterson, T. (1998). Political Roles of the Journalist. In Graber, D., D.McQuail and P.Norris (eds.), The Politics of News. The News of Politics. Washington D.C.: Congressional Quarterly (1732).
  • Pfetsch, B. (2004). Geschlossene Gesellschaft? Akteursensembles und Akteursbewertungen in Pressekommentaren. In Eilders, C., F.Neidhardt and B.Pfetsch (eds.), Die Stimme der Medien: Pressekommentare und politische Öffentlichkeit in der Bundesrepublik. Wiesbaden: VS Verlag (74105).
  • Pilon, D. (2009). Investigating Media as a Deliberative Space: Newspaper Opinions about Voting Systems in the 2007 Ontario Provincial Referendum. Canadian Political Science Review 3(3): 123.
  • Price, V. and J. Zaller (1993). Who Gets the News? Alternative Measures of News Reception and their Implications for Research. Public Opinion Quarterly 57(2): 133164.
  • Robinson, G. (1995). Making News and Manufacturing Consent: The Journalistic Narrative and Its Audience. In Glasser, T. and C.Salmon (eds.), Public Opinion and the Communication of Consent. New York: The Guilford Press (348369).
  • Rothmayr, C. and S. Hardmeier (2002). Government and Polling: Use and Impact of Polls in the Policy-Making Process in Switzerland. International Journal of Public Opinion Research 14(2): 123140.
  • Sallot, L. and E. Johnson (2006). Investigating Relationships between Journalists and Public Relations Practitioners: Working Together to Set, Frame and Build the Public Agenda, 1991-2004. Public Relations Review 32(2): 151159.
  • Schönhagen, P. (2008). Ko-Evolution von Public Relations und Journalismus: Ein erster Beitrag zu ihrer systematischen Aufarbeitung. Publizistik 53(1): 924.
  • Schudson, M. (2001). The Objectivity Norm in American Journalism. Journalism 2(2): 149170.
  • Shoemaker, P. and S. Reese (1995). Mediating the Message. Theories of Influences on Mass Media Content. White Plains: Longman.
  • Sigal, L. (1973). Reporters and Officials. The Organization and Politics of Newsmaking. Lexington: Heath and Company.
  • Singer, J. (2007). Contested Autonomy. Professional and Popular Claims on Journalistic Norms. Journalism Studies 8(1): 7995.
  • Statham, P. (2006). Political Journalism and Europeanization: Pressing Europe? European Political Communication Working Paper No. 2006/13.
  • Strömbäck, J. and D. Dimitrova (2006). Political and Media Systems Matter: A Comparison of Election News Coverage in Sweden and the United States. Harvard International Journal of Press/Politics 11(4): 131147.
  • Sturgis, P. (2003). Knowledge and Collective Preferences. A Comparison of Two Approaches to Estimating the Opinions of a Better Informed Public. Sociological Methods and Research 31(4): 453485.
  • Tarrow, S. (2004). Bridging the Quantitative-Qualitative Divide. In Brady, H. and D. Collier (eds.), Rethinking Social Inquiry: Diverse Tools, Shared Standards. Lanham: Rowman & Littlefied (171179).
  • Trechsel, A. and P. Sciarini (1998). Direct Democracy in Switzerland: Do Elites Matter? European Journal of Political Research 33(1): 99124.
  • Tresch, A. (2008). Öffentlichkeit und Sprachenvielfalt. Medienvermittelte Kommunikation zur Europapolitik in der Deutsch- und Westschweiz. Baden-Baden: Nomos.
  • Tuchman, G. (1972). Objectivity as Strategic Ritual: An Examination of Newsmen’s Notions of Objectivity. American Journal of Sociology 77(4): 660679.
  • Vatter, A. (2002). Kantonale Demokratien im Vergleich. Entstehungsgründe, Interaktionen und Wirkungen politischer Institutionen in den Schweizer Kantonen. Opladen: Leske und Budrich.
  • Watts, M., D. Domke, D. Shah and D. Fan (1999). Elite Cues and Media Bias in Presidential Campaigns: Explaining Public Perceptions of a Liberal Press. Communication Research 26(2): 144175.
  • Wuerth, A. (1999). Mediensystem und politische Kommunikation. In Klöti, U., P.Knoepfel, H.Kriesi, W.Linder and Y.Papadopoulos (eds.), Handbuch der Schweizer Politik. Zurich: NZZ Verlag (337384).
  • Yankelovich, D. (1991). Coming to Public Judgment. Making Democracy Work in a Complex World. Syracuse: Syracuse University Press.
  • Young, I.M. (1990). Justice and the Politics of Difference. Princeton: Princeton University Press.
  • Young, I.M. (2000). Inclusion and Democracy. Oxford: Oxford University Press.
  • Zaller, J. and D. Chiu (1996). Government’s Little Helper: U.S. Press Coverage of Foreign Policy Crises, 1945–1991. Political Communication 13(4): 385405.


  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Theoretical background: Criteria of fair media coverage
  5. 3. Methods and measurements
  6. 4. An overall view of press performance
  7. 5. A dynamic view of independence, inclusiveness, bias, and substance
  8. 6. A particularistic view of media fairness
  9. 7. Spatial patterns of media coverage
  10. 8. Conclusion
  11. References
  12. Appendix

Table A1: List of all ballot measures analyzed in this study

Ballot titleResult (share of yes votes)
  1. Source: Swiss Federal Chancellery.

571 10th amendment of retirement pension system (25.06.1995)60.7
572 Initiative on retirement pensions (25.06.1995)27.6
602 Law on labour: weekend and night work, maternity (01.12.1996)33.0
622 Financing of unemployment insurance (28.09.1997)49.2
643 10th amendment of pension system without increase of retirement age (27.09.1998)41.5
654 Law on labour: night work, maternity (29.11.1998)63.4
684 Law on disability insurance (13.06.1999)30.3
685 Law on maternity insurance (13.06.1999)39.0
721 Initiative against increase of retirement age for women (26.11.2000)39.5
722 Initiative for “flexible retirement age from 62 years” (26.11.2000)46.0
724 Initiative for “reduced hospital costs” (26.11.2000)17.9
732 Initiative for “cheaper drugs” (04.03.2001)30.9
752 Initiative to “secure pension system – tax energy instead of labour” (02.12.2001)22.9
762 Reduction of work time (03.03.2002)25.4
781 Initiative about gold reserves to retirement pension (22.09.2002)46.4
782 Counter-proposal about gold reserves to retirement pension (22.09.2002)46.4
792 Law on unemployment insurance: unemployment benefits (24.11.2002)56.1
802 Participation of cantons in financing of hospital treatments (09.02.2003)77.4
815 Initiative “health must remain affordable” (18.05.2003)27.1
816 Initiative “equal rights for disabled people” (18.05.2003)37.7
819 Initiative for “sufficient apprenticeship places” (18.05.2003)31.6
831 11th amendment of pension system: increase of pension age for women (16.05.2004)32.1
832 Financing of retirement pension through VAT increase (16.05.2004)31.4
844 Maternity insurance (26.09.2004)55.5