Seeing the Whole Picture. Towards a Multi-perspective Approach to News Content Diversity based on Liberal and Deliberative Models of Democracy

ABSTRACT
 By providing diverse news content, news media are key for a well-functioning public sphere. However, agreement on how to measure news content diversity is lacking. Research often refers to democratic theory as normative reference point, but different models of democracy understand news content diversity differently. Our study makes a unique, innovative contribution to this field: (1) We develop a methodological framework for measuring news content diversity, that is a set of comprehensive measuring instruments that derive different operationalizations of topic and actor diversity each from liberal and deliberative democratic theories. (2) Considering that a good public discourse requires more than diversity, we analyze news content diversity in the context of four other journalistic standards: neutrality, rationality, discursivity, and civility. (3) We prove the applicability of our measuring instruments by means of a standardized content analysis of six German news media as a case study. The different quality profiles of these outlets our analysis reveals are explained by their different functions in the media system. Our study shows how important a multi-perspectival normative approach to news content diversity is, both empirically and theoretically. Future studies on news content diversity should make their normative foundations transparent and derive their indicators purposefully therefrom.


Introduction
Healthy democracies rely on a public discourse which confronts citizens with a broad range of issues currently affecting society and a broad range of actors representing diverse opinions on how to solve current problems (Napoli 1999). This precondition enables citizens' well-informed opinion formation and political decision-making (Jandura and Friedrich 2014). By providing citizens with diverse content, the news media are key for the "proper functioning of the public sphere" (Just 2009, 98). Diversity can refer to the heterogeneity of media ownership and outlets (structural diversity), of content provided by the media (content diversity), and of content selected and perceived by the recipients (exposure diversity) (Napoli 1999). Research to date has been dominated by studies addressing news content diversity.
However, agreement is lacking on how to define and measure news content diversity (Hendrickx, Ballon, and Ranaivoson 2022;Loecherbach et al. 2020). This highly normative concept (Raeijmaekers and Maeseele 2015) describes what the media should provide, but this question will be answered differently depending on the normative foundations to which one refers. A frequent normative reference point in diversity research is "democratic theory", an umbrella term concealing various theoretical traditions. Most prominent and most commonly used are the liberal and the deliberative model of democracy (Loecherbach et al. 2020;Strömbäck 2005). However, both models understand and accentuate news content diversity differently (Helberger 2019;Raeijmaekers and Maeseele 2015), but none of them suggests evaluating news content only based on diversity. Rather, they put diversity into context with additional journalistic standards such as neutrality, rationality, civility, and discursivity, which underlines the close connection between diversity and the broader concept of media performance (Meier 2019;). Since the normative foundations strongly affect how news content diversity is measured and evaluated, it is important to make them transparent and discuss their implications in diversity research. This premise, however, has only recently been more acknowledged (Helberger 2019;Humprecht and Esser 2018). What is widely lacking are empirical studies deriving a broad range of diversity indicators systematically from different democratic models and discuss how these theories affect both measurements, findings, and interpretations (for exceptions see Beaufort 2020;Seethaler 2015). Taking such a nuanced view on news content diversity allows to analyze differentiatedly which media types that fulfill different functions in a media system make which contribution to diversity (Curran 2007;Jandura and Friedrich 2014). Our study shows how such an endeavor can be implemented. We make a unique, innovative contribution to this field by addressing three central research desiderata: (1) We develop a set of comprehensive, adequate measuring instruments (indicators and operationalizations)-a methodological framework-for measuring news content diversity by deriving various indicators of news content diversity from liberal and deliberative democratic theories. (2) Considering that a good public discourse requires more than diversity (as emphasized by both models of democracy), we connect our analysis of news content diversity with a comprehensive evaluation of other journalistic standards. (3) We prove the applicability of our methodological framework by means of case study, a content analysis of the political coverage of six German news media that we consider representatives of certain media types. By investigating the relationship between media types and news content diversity, we take an institutional approach to news content diversity (Humprecht and Esser 2018). By evaluating our findings systematically against the backdrop of the two models of democracy, we show how the measurements affect the evaluation of different news media types.
To make clear the contribution of our study, it is important to note two things: First, our case study does not aim at generalizable statements on certain media types (neither in Germany and even less beyond), which is challenging when investigating only one representative per type. Rather, we aim at developing a set of the measuring instruments which we derive from different theories and to test it empirically for the first time. We consider this a methodological framework on which other studies can build, independent of time and place of investigation. Our paper thus makes a methodological contribution. Second, taking a comprehensive, systematic approach to measuring news content diversity from the perspectives of different democratic models is a complex undertaking, and we cannot do everything at the same time. For example, there are by far more than two traditions of democratic theory (Ferree et al. 2002;Strömbäck 2005), but it would be too complex for a single paper to compare them all. Therefore, we decided to compare the two traditions fitting our case best. The German political system is a representative liberal democracy with strong deliberative mechanisms. Furthermore, our selection reflects current discussions on the representative crisis of democracy, speaking to liberal theories, and violations of discourse standards (e.g., incivility, hate speech), speaking to deliberative theories. We provide a first empirical test of how news content diversity provided by different outlets can be comprehensively compared against the background of different democratic theories. We hope that others will follow our example and complement our study by deriving indicators of news content diversity from other democratic theories and including more countries and outlets.

(Content) Diversity as a Concept
Diversity is considered a "mega concept" (McLeod & Pan, 2005) combining a variety of meanings related to the variation of media content and structure. Diversity can be broadly defined as "the extent to which media differ within themselves, in relation to certain characteristics and their distribution in society as a whole" (McQuail and van Cuilenberg 1982, 683). Mass media support society by providing structural diversity in the form of varying genres or formats, but more importantly, by showcasing a rich variety of "things that matter to people": topics, actors, geographic regions, ethnicities, and viewpoints, summarized under the term "content diversity" (Loecherbach et al. 2020).
However, the concept's definition and measurement remain fragmented (Steiner et al. 2022) due to both competing and complementary theoretical backgrounds, units of analysis, operationalizations, and mathematical measurements (Joris et al. 2020;Loecherbach et al. 2020). In two systematic literature reviews, Loecherbach et al. (2020) and Joris et al. (2020) identified three core subdimensions across which content diversity is commonly measured: (1) Actor (entity) diversity relates to individuals and groups present in media content. (2) Topic diversity assesses the issues covered. (3) Viewpoint diversity models the spectrum of positions as represented by participants' opinions. Topic and actor diversity are often measured by (predefined) lists developed for specific contexts (e.g., policy fields, candidates in national elections, different types of actors). Viewpoint diversity is mainly operationalized by examining different frames or interpretations of the same issues (Baden and Springer 2017;Steiner et al. 2022), reflecting that one topic can be viewed from different angles.
Viewpoint diversity, often considered the most important of these three, is costly to examine. Measuring it requires focusing on certain issues beyond which the findings cannot be generalized (Steiner et al. 2022). Therefore, many content analyses operationalize viewpoint diversity indirectly through actor diversity, which is much easier to measure. However, the presence of diverse actors in coverage does not guarantee viewpoint diversity since actors do not necessarily express an opinion (Baden and Springer 2017;Voakes et al. 1996). A useful middle ground (that we choose in our study) is coding only actors raising an opinion since only these participate actively in the public discourse and can disseminate their positions (Riedl, Rohrbach, and Krakovsky 2022). This reflects the normative ideal of diversity of access, requiring that different actors can contribute to the public discourse with their own opinion (McQuail 1992).
Diversity is a relational concept that must be developed and interpreted against the background of social reality. However, there is a dispute if content diversity ideally should be understood as equal visibility of all existing social entities (e.g., actors, topics) in news coverage ("open diversity"), as visibility that equals the proportion of these entities in reality ("reflective diversity"), or as a hybrid of both (McQuail 1992). While the majority of authors understands open and reflective diversity as a dichotomy, few others interpret them as two poles of a continuum (Joris et al. 2020). In any case, benchmarks are essential for evaluating diversity to answer the central question of what degree of diversity is optimal (Loecherbach et al. 2020). Despite the positive connotation of diversity, "maximal diversity is not necessarily optimal diversity" (Loecherbach et al. 2020, 624). Too high diversity can foster fragmentation, which illustrates the importance of finding the right balance between a multitude of many and a focus on certain topics, actors and opinions . But how can "optimal diversity" be assessed? This is where democratic theories come in.

Normative Perspectives on News Content Diversity
Different democratic theories assign different functions to the public sphere (Ferree et al. 2002) and news media (Christians et al. 2009;Jandura and Friedrich 2014;Strömbäck 2005). According to the liberal model, the news media should function as a hinge and an intermediary between politics and the citizenry. They should inform the citizens about societally relevant problems and solutions proposed by political actors, control the elites, make politics and the elites' actions and decisions transparent for the citizenry, enable citizens' opinion formation by presenting different positions, and articulate different interests, including political demands of the citizenry. Most studies on news content diversity and media performance build on this model (Loecherbach et al. 2020), albeit often only implicitly (Stark and Stegmann 2021). Deliberative theories, by contrast, aim primarily at engaging society into conversation and exchanging arguments in a civil manner. Social consensus should be reached based on rationality rather than on societal power structures. Therefore, deliberative theories demand a rational public discourse based on the accountability of all its participants, enabling the society to learn (Eisenegger and Udris 2021; Jandura and Friedrich 2014). (3) Relevance as measured by external benchmarks: correlation of media agenda with civil society agenda (policy issues)

External benchmarks: main topic
("what is the story about?") of … (a) parliamentary activities and press releases of parliamentary groups in the German national parliament (benchmark: parliamentary agenda) (b) press releases of civil society actors (benchmark: civil society agenda) 31 topic codes condensed into 9 broader topics (see "main topic") Actor diversity (who?) (1) Proportion between executive and legislative (1) Proportion between executive and civil society actors (2) Diversity of civil society actors: (a) Share of civil society actors present in coverage (b) Concentration of civil society actors: HHIn Actors: societal role: individual persons (e.g., politician, citizen) and collectives (e.g., political party, NGO) expressing an opinion/evaluative statement on the main topic of the article. Up to 3 actors per article; if more actors found, the three ones appearing first were coded. 146 codes condensed into 8 societal areas: domestic politics: executive branch; domestic politics: legislative branch; domestic politics: judiciary; foreign/transnational politics (incl. EU); economy; experts; civil society; citizens/public (2) Deviation from graded political equality: major parties (CDU/CSU, SPD) weighted twice; small parties (Greens, Liberal Democrats, Left, Alternative for Germany) weighted once. Analysis includes only parties represented in the German Bundestag at the time of investigation.
Actors: party affiliation (only for German political actors) 10 codes for different political parties: CDU, CSU, SPD, FDP, the Greens, the Left, AfD, Pirate Party, NPD, other parties. (1) Neutrality Neutrality: References to journalists' own opinions in news articles (opinion pieces excluded); 5-point scale from 1 (explicitly personal) to 5 (impartial-detached) (2) Rationality Rationality: Extent of emotionality within the article; 5-point scale from 1 (emotional) to 5 (rational-dispassionate) (1) Discursivity: Additive index based on four variables; index values from 0 (no aspect present) to 8 (all aspects present to a significant extent) (Jandura and Friedrich 2014;Seethaler 2015). Four other variables in relation to the main topic were coded, which indicated the extent to which it is located within a larger causal framework. They were inspired by Entman's (1993) influential definition of framing. Since this is a formative, theory-derived index (Coltman et al. 2008), its internal consistency is not considered.
Discursivity: (a) Reasons/causes, (b) consequences/possible solutions, (c) evaluations by at least two actors, (d) contextualizing information; 3-point scale for each variable (0 = not present at all; 1 = present at least to a small extent; 2 = present to a significant extent) (2) Civility: Recoding of the two civility variables into two groups: (2a) News pieces being completely civil (variable (a), code 0) or with full correction of incivility (variable (b), code 2) (2b) Uncivil news pieces (all other news pieces) Civility (a) Violations of civility (e.g., in personal attacks, discrediting insinuations, excluding moralizations, discriminatory enemy images, and ingroup/outgroup stereotypes); 3-point scale (0 = not present at all; 1 = present at least to a small extent; 2 = present to a significant extent) (b) If uncivil statements contained: journalists' handling / "correction" of incivility; 3-point scale (0 = no correction; 1 = partial correction; 2 = full correction) How these different understandings of news media's functions result in different expectations and evaluations of news content diversity (Jandura and Friedrich 2014) can be illustrated by means of three questions (Table 1; see also Eisenegger and Udris 2021;Weiß et al. 2016): what should the public discourse be about (topic diversity), who should participate in it (actor diversity), and how should it proceed (professional journalistic standards)? These questions are closely intertwined: what is discussed affects whose voices are heard, and how topics are discussed influences what citizens get out of the public discourse. Liberal and deliberative theories wanswer these questions differently (see also Strömbäck 2005).
Topic Diversity: What Should News Coverage be About?
None of these two models commits itself to any specific topics which should be discussed publicly. Rather, both call for topical openness, following the ideal of a free marketplace of ideas (Ferree et al. 2002) and advocating against a continuous concentration on individual topics. In empirical studies, this is often measured by means of diversity/concentration indices. However, none of the theories postulates topic diversity for its own sake or addressing the largest possible breadth of topics. Too much diversity can be counterproductive if it leads to a lack of focus on the most important issues . Rather, both liberal and deliberative theories demand that the news narrow down topic diversity to a certain extent to create a common social reality of problems to be solved (synchronization) . To determine how far the news fulfills this requirement to provide diversity of relevant topics, we need relevance indicators, either within the news or by applying external benchmarks. Which relevance indicators are considered suitable differs between the theories.
The liberal model considers topics with a high reach affecting (potentially) a high number of people most relevant (Eilders 2006), indicated by a focus on the macro-level of society. The deliberative model, by contrast, requires that the public debate focuses on policy issues (e.g., defense policy, social policy) rather than on politics (strategic considerations, e.g., political processes, conflict, competition). The higher the share of policy issues in all issues, the greater the chance of a rational debate based on a factual exchange about policy positions. Deliberative theories expect that this leads to enlightened rationality and increases the quality of the public discourse (Wessler 2018;Wessler and Rinke 2014).
These indicators assess relevance based on news presentation which represents the journalistic attribution of relevance. However, several authors recommend to assess the relevance of news content based on external benchmarks (Weiß et al. 2016). For example, the topics addressed in the news (media agenda) can be compared with the agendas of other societal fields. The elite-centered liberal theories demand the news media to be guided by the topics addressed in the parliament (parliamentary agenda). Deliberative theories place higher value on the news covering topics raised by civil society actors (civil society agenda).
Actor Diversity: Who Should be Visible in News Coverage?
The centrality of actor diversity in diversity research results from the assumption that only actors who are visible can raise voice publicly. More visible actors may influence the public discourse more strongly than less visible actors. Political actors more visible in the news are more likely to be considered eligible for election (Geiß and Schäfer 2017) which can lead to larger political influence. However, different models of democracy have different views on which actors should be visible in the news.
Emphasizing the representation principle and the expert norm, the liberal model focuses on political elites (including executive, legislative, judiciary, foreign political actors) and experts (Ferree et al. 2002). However, it considers the legislative-the national parliament where binding decisions for society are made-rather than the executive the heart of democracy since the strength of the political positions represented in society becomes only clear in parliaments that base on proportional representation (Althaus 2012). Therefore, the news should spend sufficient attention to the legislative compared to the executive. An indicator thereof is the proportion between executive and legislative in the news. Moreover, the news media should make the political process transparent for the citizens who mainly participate in politics by observing politics and by voting. To that aim, news coverage must both reflect the differences in size between the parties and give above-average prominence to smaller parties (Ferree et al. 2002). The principle of graded political equality combines the competing demands of open and reflective diversity (Jandura, Udris, and Eisenegger 2019). Rather than either using the proportional representation of votes or treating all parties uniformly, this principle combines (a) the requirement to make existing differences in size/importance between the parties visible (reflective diversity) with (b) the protection of minorities (open diversity) (Curran 2007;Jandura 2007).
The deliberative model, by contrast, aims at popular inclusion and equality between all participants (open diversity). Thus, actor diversity cannot end at political actors. "Inclusion of speakers from the periphery should contribute to an active dialogue between center and periphery and foster more deliberative speech." (Ferree et al. 2002, 306) Even though deliberative theories do not consider political actors unimportant, they pay greater attention to citizens as experts in their living environment and civil society actors (e.g., non-governmental organizations) bundling citizens' interests (Stark and Stegmann 2021). How far the media spend attention on civil society actors rather than only on the political center (and particularly the executive) can be measured by means of the proportion between executive and civil society actors in the news. Moreover, news coverage should not concentrate on individual civil society actors but rather provide diversity of civil society actors to ensure a meaningful public discourse and the inclusion of more voices in decision-making. In contrast to proportionality in liberal democracies, deliberative theories aim at the equal participation of all relevant actors (i.e., even small civil society groups) in the political discourse.

Professional Journalistic Standards: How Should News Coverage be Presented?
Despite the centrality of a diverse public discourse, a closer look at both democratic models shows that they by no means consider content diversity sufficient to ensure the preconditions of a well-informed citizenry. Rather, both formulate additional criteria on how the political discourse should proceed and be presented in the news (Stark and Stegmann 2021). It is nearly impossible to measure the entire range of professional standards of news quality, including for example accuracy, transparency, and attractiveness (Meier 2019), in one single study. Rather than trying to cover professional standards exhaustively, we therefore decided to illustrate how to contextualize news content diversity by professional standards by means of four standards which are particularly central to the two democratic models we focus on: neutrality and rationality for the liberal model, discursivity and civility for the deliberative model. In addition, these standards have the advantage that they can be measured much more reliably in content analyses than standards such as, e.g., accuracy and integrity. Even though these standards cannot exclusively be assigned to one theory (e.g., civility is also desirable from a liberal perspective (Ferree et al. 2002)), they represent standards that are of particular importance for the tradition we "assign" them to.
The liberal model requires a factual public discourse (Ferree et al. 2002) in which "public speakers shall refrain from emotional and polemic contributions or personal attacks, but keep the discourse purely on the ground of rational arguments and reasoning" (Jandura and Friedrich 2014, 354). This translates into rationality (Eisenegger and Udris 2021) which reflects the absence of emotions in order to ensure rational decision-making (Ferree et al. 2002) and requires a non-emotional reporting style. Furthermore, the liberal model requests that the media "inform the public in an objective and unbiased manner" (Jandura and Friedrich 2014, 355), which translates into neutrality (Eisenegger and Udris 2021) as a sub-dimension of journalistic objectivity (McQuail 1992). Neutrality refers to the journalistic norm to separate facts and opinion (Schönbach 1977) which is considered important for ensuring an independent opinion formation of the citizens (Riedl 2019). Instead of taking sides in news articles, journalists should express personal opinions or evaluations (e.g., of topics or actors) only in opinion pieces (Jandura and Friedrich 2014).
The deliberative model considers the question of how public discourse proceeds the core of its normative demands. Building on Habermas (1989), it highlights the importance of dialogue, mutual respect, and civility (Ferree et al. 2002;Helberger 2019;Wessler and Rinke 2014). This is reflected in the professional journalistic standards of discursivity and civility . Discursivity demands the justification of political claims by means of contextualizing information, mentioning reasons, discussing consequences and possible solutions, and contrasting standpoints of different actors (Habermas 1989). Elaborated operationalizations of discursivity (e.g., Delli Carpini and Keeter 1993; Seethaler 2015; Steenbergen et al. 2003) lean on the differentiation between episodic and thematic framing (Iyengar 2019) and associate thematic framing with discursivity (Jandura and Friedrich 2014). Civility means to "refrain from defamations or personal attacks that have no connection to the issue of debate" (Jandura and Friedrich 2014, 361). Different from the complete absence of emotions as required by rationality, civility allows the expression of positive emotions but rejects non-factual arguments/disputes and direct attacks/accuses which aim at excluding certain actors and their positions from the discourse (Ferree et al. 2002).

Profiles of Different Media Types
While democratic theories address their expectations to the public sphere and news media in general, "there are different ways to 'perform' and […] different kinds of news media contribute to the overall news environment in a specific way" (Hasebrink and Hölig 2020, 299). Even though we cannot assign certain media types to certain models of democracy, different types of media fulfill different roles in a media system. This can make it easier or harder for them to meet the expectations of different models of democracy (Jandura and Friedrich 2014). This must be considered when comparing news content diversity provided by different media types but has so far been undertaken too little. So far, studies on news content diversity often compared only quality newspapers vs tabloids (e.g., Magin and Stark 2011) respectively PSB vs commercial television (e.g., Donsbach and Büttner 2005).
In our study, we focus on six outlets representing particularly important political information sources in Germany (Hölig and Hasebrink 2021). Table 2 gives an overview of them. Certain characteristics of these media types have implications for the news content they can provide. For example, TV newscasts are quite short (Tagesschau 15 minutes, RTL Aktuell 20 minutes), thus offering only limited space for diverse information and discursive formats. Precisely because of this brevity, it is important that TV newscasts pay particular attention to diversity. For discursivity, the far more comprehensive newspapers and especially news magazines offer much better opportunities. Differences like that must be taken into account when later interpreting our findings.

Research Questions
Our goal is to develop and empirically test a methodological framework consisting of a set of indicators and operationalizations that enables to compare news content against the background of liberal and deliberative democratic theories. From the previous considerations, we derive two research questions: RQ1. To what degree do which news media types in Germany meet the normative expectations of liberal democratic theories concerning topic diversity (RQ1a), actor diversity (RQ1b), and professional journalistic standards (RQ1c)?
RQ2. To what degree do which news media types in Germany meet the normative expectations of deliberative democratic theories concerning topic diversity (RQ2a), actor diversity (RQ2b), and professional journalistic standards (RQ2c)?

Sample and Data Collection
We answer these research questions by means of a quantitative content analysis of news coverage about domestic politics, a field in which both liberal and deliberative theories consider comprehensive information an important precondition for a well-informed citizenry (Jandura and Friedrich 2014). We compare representatives of the six media types described above: a quality newspaper (FAZ), a tabloid (BILD), a regional newspaper (Rheinische Post), a public service newscast (Tagesschau), a commercial TV newscast (RTL Aktuell), and a news magazine (Spiegel). We selected these individual outlets since they are the most used outlets of their type at the national level (BILD, Tagesschau, RTL Aktuell, Spiegel) or are widely distributed representatives of media types that are central to the German media landscape (FAZ, Rheinische Post). We analyze the offline versions of these outlets which are still more important as news sources than online outlets Broad range of issues and actors Public service broadcasting (PSB) (predominantly) publicly financed in order to protect it from market pressure and to ensure that it can fulfill its obligations (Lund and Lowe 2015) Commercialized media (tabloids, commercial TV news) Making profit-but also bringing issues neglected by other media types (Örnebring and Jönsson 2004) Largest possible audience Rather unpolitical focus, but potential to take up issues neglected by other media types (Örnebring and Jönsson 2004) Strongly dependent on advertising revenues, therefore strong pressure to popularize content

Regional newspapers
Taking up issues relevant to the distribution area neglected by other media types (Hess and Waller 2017) Population in the distribution area Regional issues National coverage often produced jointly for several regional newspapers belonging to the same publishing house News magazines Providing "in-depth articles and interpretive features on recent events" (Scott 2008, 1) and comprehensive context information Broader public Recent events Susceptible to popularization as measured by the use of colors, mixing hard and soft news, incorporating emotions (Umbricht and Esser 2016) and social media in Germany (Hölig and Hasebrink 2021). An additional online-offline comparison would be beyond the scope of this already quite complex endeavor. As case, we chose Germany, a typical example of a democratic-corporatist media system with a high newspaper circulation and strong position of PSB (Hallin and Mancini 2004). Due to its structural similarities with many other European countries, Germany represents a meaningful case for analyzing news content diversity. The number of German regional newspapers far exceeds that of national newspapers and news magazines. There is an ideologically diverse segment of five national quality newspapers with clear editorial lines, including the conservative FAZ (Thomaß and Horz 2021). The only tabloid being published nationwide is the most-read German newspaper BILD. Among the few national news magazines, the left-wing Spiegel is most prevalent (Hölig and Hasebrink 2021). On the broadcasting sector, PSB (financed by licensing fees) and commercial providers have been competing since the 1980s (Thomaß and Horz 2021). The currently most widely used newscasts are the Tagesschau (PSB) and RTL Aktuell (commercial) (Hölig and Hasebrink 2021).
Our dataset stems from an internationally comparative study which collected data in three countries during eight weeks in 2018, 1 a year without national elections in Germany. Sampling was done in three steps: (1) Through technical pre-filtering, 2 we identified potentially relevant news items (including both news articles and opinion pieces) in all news items on domestic politics published during the investigation period. Our broad definition of domestic politics comprises national politics, regional politics, foreign affairs (the latter two only when referring to national politics), and issues of societal interest which are negotiated in national political institutions (e.g., unemployment statistics, food pollution caused by new pesticides). Our search string for identifying political news goes beyond usual procedures in content analysis on political news coverage that select only articles published in the politics section or containing party names. Rather, we searched not only for the names (abbreviations) of the most important political parties in Germany (e.g., "CDU", "SPD") but also for those of international organizations/associations (e.g., "UN", "G7") and pivotal genuine political terms also beyond the political center (e.g., "parliament", "demonstration", "citizen", "movement"). For political terms not exclusively linked to national politics, we additionally required the appearance of the word "German" (for the full search string see Appendix).
(2) For all identified news items (about 30% of all articles), three student coders decided on inclusion in the sample. Following the assumption that news items typically start with the most important aspects, the decision on inclusion based for written text on the articles' heading and lead or (if no lead available) first paragraph, for audio-visual pieces on the introduction by the moderator or (if no introduction available) the first 20 seconds. The student coders agreed satisfactorily on the identification on relevant news pieces (n = 170 articles; agreement: 77-95%; mean: 89%).
(3) For outlets with a comparably low number of news items identified in step 2 (Spiegel, n = 149; Tagesschau, n = 193), we coded all these news items. For all other outlets, we drew a random sample (sampling error: 3%) from all news items identified in step 2 (FAZ, n = 618; BILD, n = 264; Rheinische Post, n = 256; RTL Aktuell, n = 183). The final sample includes 1663 news items.
As external benchmarks for the relevance of news content, we coded the topics of 1415 parliamentary activities and press releases of parliamentary groups in the German national parliament and the topics of 778 press releases from 104 civil society actors, using the same topic categories as for the news articles (see below and Table 1). Due to the lack of a systematic register of civil society stakeholder groups in Germany, we selected 29 civil society actors representing a broad spectrum of societal areas/interests, including both society-wide (e.g., churches, trade unions) and special interests (e.g., sports, environment). Table 1 provides an overview of the categories we coded and how we transformed them into different indicators for our analysis. Both the main topic and up to three actors raising positions were coded manually per article. Compared to automated analyses (e.g., Vogler, Udris, and Eisenegger 2020) which must focus on written text and recognizing the names of actors, manual coding has the advantage of allowing to code both textual and audiovisual material (TV news) and to only code actors expressing a position, which relates actor diversity closer to viewpoint diversity, as described above. Coding up to three actors per article might have contributed to a strong focus on actors from the political center, affecting our results on actor diversity. The inverted pyramid used in journalism requires to place the most relevant at the beginning of a news item, and political actors are undoubtedly relevant in political reporting. However, if journalists often place political actors first, this still says something about the elite-oriented self-understanding which many of them obviously feel committed to.

Measurements
To give the coders the best possible orientation how to code topics and actors, we used detailed lists for coding the articles' main topic and the actors' societal role. For our analyses, we aggregated these differentiated codings to more general categories. Detailed category descriptions and how they were recoded can be found in the Appendix.
The aggregation of the topics (from 31 topics to 9 policy fields) followed first the established differentiation (Vowe 2008) between polity (here: political structures and culture), politics (here: political actors), and policy. Second, we assigned the policy issues to seven policy fields representing both central societal areas and German federal ministries: internal security and law/justice, foreign policy/international politics, economy/finance, social policies, culture/media/education/sports, migration, and agriculture/environment/energy. Even though migration does not have its own ministry, we kept migration as a separate policy issue because of its particular prominence in political reality during the investigation period.
The aggregation of the actors (from 146 roles to 8 societal areas) is based on the juxtaposition of political elites (center) and civil society actors (periphery) (Wessler and Rinke 2014). The eight societal areas used in our analyses contrasts four political elite groups representing the political center (executive (head of state and national government), legislative (national parliament and national parties), judicial (national courts and judges), foreign/transnational politicians (Humprecht and Esser 2018;Masini et al. 2018;Benson 2009)) with four other societal areas (economy (which we consider an important separate entity), experts, civil society actors, and citizens (e.g., victims, volunteers, strikers, protesters; Beckers and Van Aelst 2019)). The expert category includes professionals from different fields (media and journalists, business professionals, academics). Following Beckers and Van Aelst (2019), we included also celebrities who are in the German media (e.g., in political talk shows) often portrayed as experts/professionals in their field/lifeworld, contribute to the discourse by expressing an opinion on the respective issue, and are in the news clearly distinguished from citizens. Civil society actors represent social organizations/interest groups oriented to various issues (e.g., health policy: representatives of doctors, patient organizations … ; housing policy: representatives of owners and tenants …) whose importance is strongly anchored in the deliberative model. In order to provide an inclusive picture of civil society actors in Germany, we captured both strongly institutionalized organizations (e.g., unions, churches, political movements) and representatives of issue-specific civil society actors.
As index for topic and diversity of civil society actors, we calculated the normalized Herfindahl-Hirschmann-Index (HHIn) which is common in media economics to describe concentration processes in markets (Just 2020). While Shannon's H-which is popular in diversity research-measures diversity (Mcdonald and Dimmick 2003), the HHIn measures the absence of concentration (and thus the presence of diversity). Both indices take "as benchmark the notion of open diversity [aiming] at flattening the distribution, making the different elements equal" (Loecherbach et al. 2020, 621). Empirical studies show similarities between the HHIn and Shanonn's H, indicating that both can be treated equivalent (Loecherbach et al. 2020). Compared to Shannon's H, however, the HHIn has the advantage of fixed value ranges indicating no (below .15; diverse reporting), medium (.15 to .25; moderately concentrated reporting), or high concentration (above. 25; highly concentrated reporting) which are helpful for interpreting our findings (Naldi and Flamini 2014, 3). Moreover, different from Shannon's H, the normalization of HHIn results in a measure that is comparable to other concentration calculations, even if the number of topic/actor codes differs from ours.
Neutrality, rationality, and discursivity were coded on different scales. To make them comparable, we z-transformed them by subtracting the mean of the respective scale from the original value and dividing it by the standard deviation. The resulting standard unit represents "how many standard deviations an observation moves away from the mean, as well as the direction in which it moves" (Banas 2017(Banas , 1901.

Reliability
The 1663 news items were coded manually by 14 student coders. We tested for reliability by means of Brennan and Prediger's kappa, which is chance-corrected and more robust than Krippendorff's alpha regarding variables with a skewed distribution as ours (Quarfoot and Levine 2016). The reliability coefficients for topics and actors were calculated after aggregating the codes into the overarching categories. Reliability (n = 80 articles) was satisfying for all variables (topic (media agenda): .69; actor role/function: .92; party of political actors: .93; societal level: .68; violations of civility: .77; journalists' handling of incivility: .76; neutrality: .80; rationality: .76; discourse index: .71). Topics of parliamentary activities and press releases were manually coded by six student coders with good reliability (n = 38 documents; topic (parliamentary agenda & civil society agenda): .74).

Findings
Topic Diversity: What is News Coverage About? (rq1a/2a) Topic diversity (HHIn). Both the liberal and the deliberative model value a high diversity of issues. Our findings show that all six outlets meet this criterion, indicated by HHIn values smaller than .06. Even though all outlets show non-concentrated reporting, there are notable differences (Table 3): The quality paper reports most diversely, followed by the regional paper, the PBS, the commercial newscast, the tabloid, and the news magazine.
Relevance within coverage. Concerning the share of articles focusing on the societal macro-level (Figure 1) which is most important from the perspective of liberal theories, we find again clear differences between the outlets. The quality paper, PBS newscast, and regional newspaper fulfill this criterion to a significantly higher degree than the commercial newscast, news magazine, and tabloid. The deliberative model, by contrast, calls for a stronger reference to policy issues compared to politics or polity. When adding the seven policy issues in Table 3, the quality paper comes closest to this criterion (74.8% policy topics; sum of policy issues not displayed in Table 3), followed by the PBS (70.6%) and commercial newscast (65.7%), regional paper (64.1%), tabloid (62.4%), and news magazine (56.2%).
Relevance as measured by external benchmarks. For investigating how strongly the media reflect the policy issues on the parliamentary agenda (as a criterion for the liberal model) and the civil society agenda (as a criterion for the deliberative model), we determine correlations between the topic rankings of each outlet (excluding political actors as well as political structures and culture) with the two other agendas (Table 3; for an overview of all three agendas see Table A1 in the Appendix). Correlations with the parliamentary agenda are strongest in case of the quality paper, news magazine, and regional newspaper, which show positive correlation coefficients with the parliamentary agenda. We found no or negative correlations between media coverage and the agenda  Note. χ2(40) = 110.41; 1 weighted according to the ratio between sample size and population for each outlet; 2 31 coded single topics condensed to 9 topic groups as shown in the table; 3 bivariate correlations based on topic shares of the 7 policy topics; 4 parlamentary activities and press releases. 5 Aggregated mean values across media outlets. Since migration was the dominating political topic at the time of investigation, we decided to use it as separate topic.
of civil society actors. The ranking is led by the regional newspaper, commercial newscast, and quality newspaper.
Actor Diversity: Who is Visible in News Coverage? (rq1b/2b) Proportion between executive and legislative. Articles covering political actors (including executive, legislative, judiciary, and foreign political actors) and experts represent the largest share in all outlets studied (Table 4), as required by liberal theories. However, there are differences between the outlets: in the tabloid, political actors and experts are covered in 87.7% of the news items (sum of A1, A2, A3, A4, and A6; results not displayed in Table 4), followed by the news magazine (85.5%), PBS newscast (82.1%), regional newspaper (81.6%), quality newspaper (77.0%), and commercial newscast (76.8%). However, problematic from the perspective of liberal theories is the overweight from the executive over the legislative as the heart of democracy (Althaus 2012). The smallest executive overhang (A1 divided by A2; results not displayed in Table 4) is found in regional newspaper (2.1 times more articles with executive actors), followed by PBS newscast (2.4), news magazine (2.6), quality newspaper (2.8), tabloid (3.1), and commercial newscast (4.3). Deviation from principle of graded political equality. The liberal model demands graded political equality in news coverage. We investigate how far this is fulfilled by calculating the deviation from this principle for each possible combination of parties represented in the German Bundestag at the time of investigation in 2018 (e.g., SPD vs FDP, SPD vs Left). According to this principle, the major parties CDU/CSU and SPD (which together held 56% of seats at that time) should only appear twice as often as the smaller parties FDP, Greens, Left Party, and AfD. The news magazine comes closest to this criterion with an average deviation of 2 percentage points per party constellation, followed by the PBS newscast, regional and quality newspaper, commercial newscast, and tabloid (Table 5).
Proportion between executive and civil society actors. Deliberative theories place special emphasis on the representation of civil society actors in the news. These occur most frequently in the quality newspaper, followed by the regional newspaper, PBS newscast, news magazine, tabloid, and commercial newscast (Table 4). To ensure a sufficient consideration of civil society interests in the public discourse, deliberative theories require, moreover, that civil society actors should not be neglected in relation to executive actors. However, none of the six outlets meets this requirement (A1 divided by A7; results not displayed in Table 4). The dominance of the executive over civil society actors and citizens is lowest in the quality newspaper (3.9 times more executive actors), followed by the regional newspaper (5.1), PBS newscast (5.7), news magazine (9.4), tabloid (14.1), and commercial newscast (14.7).
Professional Journalistic Standards: How is News Coverage Presented? (rq1c/2c) Neutrality. Concerning the question of how news should be presented, liberal theories demand that journalists do not refer to their own opinions in news items but only in opinion pieces. An analysis restricted to the news items in our sample shows that the norm to avoid personal viewpoints in news is best met by the regional newspaper, the PBS newscast, and-with some distance-the quality newspaper. It is less well fulfilled by the commercial newscast, the tabloid, and particularly the news magazine ( Figure 2).
Rationality. A similar picture emerges concerning rationality, the second professional standard derived from liberal theories. We find the most rational presentation of news in the regional newspaper, PBS newscast, quality newspaper, and commercial newscast, while the news magazine and tabloid report least rational (Figure 2).
Discursivity. Concerning discursivity, we find that the news magazine, quality paper, and PBS newscast are most in line with this pivotal demand of the deliberative model. The regional newspaper performs significantly lower, but furthest from the ideal are the commercial newscast and particularly the tabloid (Figure 2). Figure 2. Professional standards of reporting (z-standardized): Neutrality, rationality, discursivity. Note: Neutrality: n = 1461, rationality: n=1663, discursivity: n=1663 news items; means as basis for z-standardization weighted according to the ratio between sample size and population; *basis: all articles except commentaries, columns and editorials.
Civility. Civility as required by deliberative theories is overall very high across almost all news media under investigation ( Figure 3). All outlets come quite close to the ideal of civility; even the news magazine as only outlet providing a significantly lower degree of civility presents most pieces content in a civil manner.

Discussion
The study at hand is pioneering in addressing several research desiderata that have long been pointed out in diversity research (e.g., Loecherbach et al. 2020). It is the first one that systematically derives a broad variety of operationalizations for news content diversity and professional journalistic standards from two normative models of democracy. This challenging, complex endeavor has frequently been called for in the literature but hardly been implemented to date (Jandura and Friedrich 2014). Our comprehensive, multi-perspectival approach shows that such an endeavor can be realized in a beneficial way and is worth the effort: it leads to a differentiated picture of news content diversity that opens up for broader comparisons across outlets, media types, and countries and their evaluation against the backdrop of different democratic theories.

Democratic Theory Affects Selection of Indicators
As our study illustrates, many established variables for measuring news content diversity and media performance can be utilized for operationalizing the demands of different models of democracy by using them for a variety of analyses. However, one must be aware that the democratic model chosen strongly affects the selection of indicators and therewith the evaluation of news content. For example, the news magazine Spiegel "performs" overall somewhat better from a liberal than from a deliberative perspective, while this is just the other way around for the commercial TV newscast RTL Aktuell. Even more surprising may be that the assessment of individual media types often differs less between indicators derived from different democratic models than between Figure 3. Professional standards of reporting: civility. Note: n = 1633 news items; χ 2(5) =47.356; p<.000; *weighted according to the ratio between sample size and population for each outlet. different indicators derived from the same model. For example, we find a discrepancy between relevance within coverage and as measured by external benchmarks for some (but not all) outlets: The PBS newscast Tagesschau shows a relatively high internal relevance but a rather low external diversity, as measured by both the liberal and deliberative indicators. The pattern is reverse for the news magazine Spiegel. The regional newspaper Rheinische Post shows both a medium internal and external relevance from a liberal perspective compared to a rather low internal relevance but high external relevance from a deliberative perspective. How strongly the indicators affect the findings calls for choosing both democratic theories and indicators wisely. But these findings show also how important differentiated measuring instruments are to draw an accurate picture of the outlets' performance. It is clearly advisable to rely on more than one indicator per dimension of news content diversity, even if using only one democratic model as theoretical background.
Our first empirical test has proven the applicability of our methodological framework. Still, the set of indicators and operationalizations that we suggested is not exhaustive. For example, we neglected geographical diversity or demographic diversity of actors (e.g., gender, ethnicity). Our approach requires, moreover, further systematic work both theoretically and empirically in order to better understand the advantages, disadvantages, and interaction of the various measurements and indicators. To that aim, cross-national and long-term comparisons are advisable. A pioneer study such as ours may not be able to compare countries right away. However, investigating Germany as an established democracy has proven our methodological framework to be fruitful. In the next step, our indicators and operationalizations should be validated by transferring them to other contexts. We consider it important to compare news content diversity also beyond the Western context (e.g., in establishing, declining, illiberal democracies and autocratic systems) and eventually develop more specific indicators to that aim. Our study provides useful analytical tools for such comparisons.

Media Types and Outlets Affect Performance
Our case study of six German news outlets provides new insights into how far these meet the requirements that different democratic theories place on the public sphere. However, "the public sphere" is composed of different media types with different functions, which these theories neglect. For example, Spiegel is highly discursive but performs worse at neutrality, rationality, civility, and topic diversity. This may be criticized from a liberal perspective but shows in fact that Spiegel uses its extensive coverage to fulfill its function as a news magazine: it provides in-depth reporting on certain core issues (rather than news on a broad range of issues). The PBS newscast Tagesschau provides highly neutral, rational news but scores lower on discursivity. This might be criticized from a deliberative perspective but is perfectly in line with the specific function of the Tagesschau to quickly update large parts of the German population on the latest, most important events and topics. These clearly distinct profiles of the different media types may to a large extent also be explained by their orientation towards the (presumed) interests of their main target groups which strongly differ between, for example, TV newscasts, tabloids, and quality newspapers (Riedl and Eberl 2022).
Despite the proximity of some outlets to certain models of democracy, our findings show that none of them completely implements the ideals of any theory. Not even the quality paper FAZ, even though this media type is often considered a "benchmark" for evaluating news content from a liberal perspective on democracy. Democratic theories describe ideal-typical expectations which hardly ever can be fully met in reality. As our study clearly shows, a realistic evaluation of what individual outlets contribute to which model of the public sphere requires more than only deriving indicators from normative democratic theories. Rather, we must also consider the actual functions and format of media types which affect the performance they realistically can achieve. A media system only consisting of quality newspapers that meet the needs of an elite readership is not desirable. It would neglect the needs of other segments of the citizenry who much more strongly rely on other media types. From the users' perspective, a structurally diverse media system, comprising a variety of media types, is preferable.
How the functions of different media types are fulfilled can, however, also differ between different outlets belonging to the same type. News content is clearly influenced by the conditions under which individual outlets are produced (e.g., newsroom size and structure, journalists' working conditions). For example, the regional newspaper Rheinische Post might score high on rationality and neutrality because smaller newspapers often rely on newswire copy ) and do not have the financial means to invest in original (more opinion-oriented) reporting. The six individual outlets we investigated do by no means speak for all other outlets of their types. Future studies should complement content analyses on news content diversity with investigations of individual outlets' production conditions. Worthwhile would also be a comparison between the outlets' offline and online versions.

Democratic Theory Affects Evaluations
In our study, we compared the liberal and the deliberative model of democracy which fit the political reality of Germany as our case best. This has naturally affected our findings and evaluations, best illustrated by the comparably poor performance of the tabloid BILD and the commercial TV newscast RTL Aktuell. These media types have functions in the German media system that are not considered important by liberal and deliberative theories.
However, there are other democratic theories (Ferree et al. 2002;Strömbäck 2005), and performing poorly in the eyes of one theory can indicate a good performance from the perspective of another theory. For evaluating the contribution of commercial, popular media to the public sphere (Örnebring and Jönsson 2004), participatory theories may be a more appropriate yardstick which future studies might add to our measuring instrument. These theories have a broader understanding of politics and aim more strongly at participation, popular inclusion, and empowerment of citizens (Barber 2009;Jandura and Friedrich 2014). To give just one example how the theoretical perspective can affect the interpretation of findings: liberal theories criticize a lack of rationality (i.e., using emotions), which runs counter their ideals of the public sphere (Jandura and Friedrich 2014). Participatory theories, by contrast, consider the use of emotions positive if it helps to empower citizens and bring them closer to politics.