Introduction

As a continuous phenomenon in the relationship between news media and political institutions, party bias is one of the most thoroughly researched practices in the field of political communication, organized civil society and electoral authorities in regards to news media performance. During elections, the most commonly used procedure consists in carrying a quantitative content analysis of news pieces published by a sample of media outlets. Nevertheless, such procedure faces methodological problems that ignore certain aspects of performance. Two examples could be pertinent: firstly , the final or full balance of media coverage demonstrate an impartial behavior from news media, with further qualitative or linguistic analysis showing some forms of bias that are not registered by the quantitative procedure, as the use of passive voice , ironies, or impersonal sentences that gain, more significance (^{Salgado, 2009}). A second issue is big samples nullify some highly visible and publicly significant events in the end result, such as presidential debates. The final quantitative balance masks the fact that these media outlets, in some particular moments, skew the coverage to be favorable for either party.

Both cases reflect content analysis has some blank points that hinder its accuracy from the beginning, a common limitation of the techniques of the positivist paradigm. Thence, we assert that such issues are not linked to this problem, but to other standardized procedures of measurement that are applied without considering their methodological limitations and their lack of auxiliary procedures. Therefore, this paper aims to examine the problems of the standardized procedure of content analysis, at least the ones cited in the national and international literature reviewed, to acquire validity in the measurement of party bias; and we suggest some strategies such as additional indicators and sample procedures to face them. We present an example of the application of these proposals, the coverage of the Mexican elections of 2012 that pretends to demonstrate its efficacy, its scope and limitations. Hence, we would like to invite scholars to critically review their procedures in order to improve them, away from the inertias and on which these have rooted in the institutions that observe election news.

The party neutrality aspiration. Normative fundamentals

In the political field of journalism, bias is defined as a systematic tendency to favor some actors or positions to the detriment of another, by selecting a source that produces incomplete information about subjects entitled to them, the uneven access dimension, or through favorable or unfavorable editorial treatments, the uneven treatment dimension (^{Fico, Freedman, & Love, 2006}; ^{McQuail, 1998}). Such practice has a significant political dimension since it's a transgression of the norms of balance and objectivity in journalism performance, i.e., the presentation of different aspects of an issue or event, neutrality in tone and form, emotional restraint, separation between facts and opinion, and absence of a double agenda (^{Christians, Glasser, McQuail, & White, 2009}; ^{McQuail, 1998}; ^{Schudson, 2001}).

Transgression of both principles have deep political consequences for democracy , since the equal representation of political options in news media is a necessary condition to seek the equal right to be elected and the competitiveness of the elections. This assumption is essential considering the advantages in resource control opinions the elite groups have to express themselves in public. This could reduce the alternate sources of information and eventually hinder the chances of the less powerful political forces to compete (^{Cantú, 2013}; ^{Dahl, 2000}). On the other hand, from a normative point of view, balance or impartiality seeks to attend the principle of diversity by which media are thought as public spaces that should evenly "reflect, the plurality of political and social groups and interests, in accordance with the ‘public interest’ (^{McQuail, 1998}). Both stances strive to correct structural inequalities in the political systems that hinder democracies.

These elements are the basic principles of normative assumptions that inform the assessment of journalistic performance, carried by several actors such as the electoral authority, organized civil society and academia; all of these have developed methodological procedures to translate those assumptions to concrete, objectively verifiable observations. In the meantime, they have acquired some issues that stem from the conceptual inadequacy of said principles to the media systems assessed, or an incomplete operationalization of party bias. In the next section, we explore both aspects and propose some practices that could help fix these problems.

Academic measurement of party bias. Critical observations

From a thorough review of the literature of party bias, ^{Niven (2003)} finds that a homogeneous, taken for granted, principle underlies it, and it is implicit in the analysis of the data that are used to conclude the existence and reach of party bias. This author proposes de term call "proportionality rule": in a political system with two or more options that produces an equal number of candidates, equally qualified, and every campaign contexts events, activities, and discussion almost in equal quantities. Then, the coverage should correspond to that proportion, and every deviation from that rule could be considered a consequence of some sort of political affinity.

Nevertheless, this assessment criteria doesn´t hold up to the actual conditions in which the media act, making those preferences unrealistic or even undesirable to fulfill, and present a way of measurement that has significant limitations (^{Niven, 2003}). We propose three kinds of them, as a manner of objections to that expectation: firstly, a historic cultural one, which considers that party bias in some media systems, is part of a journalistic tradition and not a transgression. Secondly, an organizational one, that is, the differentiated abilities of political actors to fit into media routines; and a methodological one, in how bias is operationalized and measured. These three conditions compromise the validity of bias assessments when those are measured and interpreted as an expression of an editorial effort to favor some parties instead of another, without noticing those factors.

The opposite to party bias, that is balance, is a professional norm that stems from the "objective" model of journalism, an Anglo-centric professional ideology whose contingency is interpreted as the dominant, central and sometimes essential expectations in many western media systems (^{Hallin & Mancini, 2004}). In practice, in some of those societies, this ideology is not sustained, partly because of the historical inhibition of the ethos of public interest (^{Papathanassopoulos, 2001}); nonetheless, their promoters in academia, government, and the journalistic field uphold that expectation, without considering the endogenous circumstances that constraint and, thus, moderate its likeliness.

From another perspective, partiality in those systems is a pretty conventional and even desirable trait: it happens in the advocacy role of some "committed" and openly ideological and partisan press; In the facilitator role, that in the name of the principles of social responsibility and the practice of journalism as a means to improve the way of life of societies, is entitled to be biased to comply with some understanding of public interest (^{Christians et al., 2009}). In the polarized plural systems, where the diversity of rival groups parallels a heavily biased media landscape that caters to their own niche markets, known as external diversity (^{McQuail, 1998}). In these cases, journalists behave according to a cultural logic akin to party bias, because it is expected they will commit to report a particular selection of events and offer a self-referential interpretation of reality, however, they should sustain the principles of journalistic professionalism. In that logic, , being neutral is interpreted as a lack of commitment, civility or political allegiances, such as the cases of a Mexican journalistic tradition of denunciation or the journalistic cultures of Holland and France (^{Hallin y Mancini, 2004}).

The contingent position of journalism in relation to social expectations makes impossible to be definitive about the "universal" norms that should apply everywhere (^{Christians et al., 2009}). Many works, however, suggest a research design and data interpretation that stems from the normative assumption that a neutral or objective media performance in elections coverage it´s an essential component of a democratic system; and that party partiality is a dysfunction of it, more unsettling the deeper and generalized it is. The circumstances we previously described contradict such stance, because the assessment of bias depends on the real conditions of a given media system.

The second source of bias, now organizational, comes from the merits, abilities, professionalism, and performance of certain spin-doctors or campaigns, to adapt to media news-making processes, in detriment to other teams. Because of the limited resources of the media to "report reality", they take the supplies of events and information that campaigns provide, though in different qualities and quantities. Thus, a well-adapted campaign to the news-making process could organize flashy events, ready to report; fit their communications to the language and values of the news media; and present more newsworthy, prominent and citable candidates, based on their traits of "media personality": credibility, affectivity and richness of ideas and experiences (^{Fico et al., 2006}; ^{Niven, 2003}; ^{Shoemaker & Reese, 1996}). In theory, a better treatment of the media to such candidate could make them look biased, but in fact he is more resourceful to prompt an affordable, attractive and, thus, favorable coverage; in fact, there would be bias if the media produces equal coverage even though some actors successfully adapt to their routines, and others remain passive, unaffordable and unprofessional.

At the same time, the real unevenness of the party system could benefit a party instead of another; it's a given that there exist differences in the political weight of parties depending on the size of their constituencies and their roles in the political system (majority, minorities, issue specific). This causes, uneven levels of newsworthiness due to their norms of hierarchy (^{Hopmann, Van Aelst, & Legnante, 2012}). Likewise, some sorts of events attract more coverage than others, such as errors, gaffes, public demonstrations, etc. Hence it can be possible that bias towards any side reflects power unbalances and the fickle nature of campaign events, that ends up tilt the balance towards them. Also, the assessment that does not set a difference between coverage differences that come from party bias and those who come from the better performance of a given campaign that has the abilities and adapts to the news making process, could be spurious (^{Schiffer, 2006}).

The last problem is about the epistemic and methodological assumptions and procedures commonly used to measure bias, from which we underline three of them. Perhaps the most basic and complex at the time is the "tone" category of the news items or valence, in many works the only indicator of editorial treatment, typically classified as positive or negative from the adjectives used in the pieces. This is due to the operationalization procedure which identifies explicit bias, ignoring those latent forms that use more sophisticated discourse strategies as the ones mentioned in the introduction (^{Salgado, 2009}). On the other hand, many works do not report reliability measures for said variable, and when they do, reliability is not quite achieved (^{Fico, Zeldes & Diddi, 2004}).

Another methodological issue is that every news item or sample unit has the same quantitative value that the rest of the pieces of the corpus. In terms of concluding inferences, by aggregating cases, the significance and potential impact of each of the events covered -of extreme variation- is equated, so that the editorial policy that precedes the treatment, whether biased or neutral, could seem consistent for all the events covered, giving "the same weight to different events” (^{Graber & Dunaway, 2015, p. 353}). Indeed, from time to time some media outlets could exert biases that, even sporadic and quantitatively insignificant, seek to have a significant impact on audiences. That could be the case of televised debates. Knowing that such events draw a very large audience, an outlet could give a party higher visibility and better treatment, but that bias is compensated afterward with a balanced coverage of the trivial campaign events of the other party. By not pondering the public impact of the event, the quantitative analysis will conclude that this coverage is fair.

In conclusion, these studies are problematic. The Content analysis method was conceived, from its inception, with a twofold purpose: to describe and to interpret data, inasmuch it allows to infer the intentions of the sources (^{Riffe, Lacy & Fico, 1998}). The problem is that many works interpret the data in a decontextualized manner, and an "aligned" editorial policy is concluded, without contrast them to actual regulation about media performance, that should be taken into account, or the ability of any candidate to successfully adapt to the news making process and thus achieve better coverage (^{Groeling, 2008}).

Some methodological proposals to the problems discussed

To solve these discussed issues, some logical and literature based solutions are proposed. Firstly, to carry out more sophisticated measurements that complete the dominant measure of "tone" or "valence", which could catch up with the increasing sophistication of the media outlets to produce bias. One of them could be the development of an index bias that sums structural indicators about the visibility of the candidates: the order by which they appear, their time or space of allocation, direct quoting, etc. (^{Fico et al., 2004}); or to identify a convergence of the agenda between the news media and any party (^{Graber & Dunaway, 2015}). Another way to sophisticate the measurement is to carry longitudinal studies that show patterns of consistency in the media system (^{Hopmann et al., 2012}) since the crosscutting measurements could be influenced by very professional, media friendly campaigns that could entail structural biases rather than partisan´s.

On the other hand, it would be possible to combine census or samples about journalistic coverage with "base line" procedures; that is, events or conditions more or less objective, of a consensual cultural interpretation -like homicide rates-so that a given coverage could be expected with confidence (^{Schiffer, 2006}). If any distortion of the treatment occurs; it is possible to attribute it to party bias rather than the news making process.

Likewise, it is recommended to consider two ideas about the conclusions on bias behavior. Firstly, to expect or interpret a media performance based on the endogenous journalistic traditions and media system on which they act. That would entail, for example, to consider as satisfactory a final result of external diversity -bias per outlet, balance on the whole market- rather than privilege internal diversity -balance within each outlet- as the prime criteria of editorial performance. A media system assessment, with their own traits, traditions, and contradictions, rather an individual analysis, is recommended to reach that goal.

Secondly, it is necessary to take into account the issue of the symbolic weight of the parties in the public space and the merits of the performance of some campaigns above others, emphasizing the abilities, professionalism and public relation might of the parties, and de-emphasize the influence of the editorial policy in the final coverage of a campaign. In that case, it would be possible to establish party bias when the data show that media do not use news-worthiness criteria to cover politics, such as the stance and relevance of parties and actors, inferred by the proportions of seats in Congress or the outcomes of the elections. Also, their position on the poles, that makes them salient in a horse-race coverage frame- charisma and popularity of the candidates, or their focus on conflict events rather consensual ^{Hopmann et al., 2012}). If these patterns are present in a proportional way amongst the parties, but the coverage is clearly favorable to one of them, then it would be possible to conclude that a party bias is occurring.

Such assessment entails a scientific practice that makes cautious and multifactorial inferences and tries to disentangle bias that arises from the nature of journalistic field -structural- from those that come from ideological, partisan or clientelist factors.

Implementing the solutions. Brief methodological test on the Mexican case

The next study seeks to demonstrate how convenient is to execute methodological strategies that attenuate the issues previously discussed while testing, some of the solutions proposed. The context chosen to do so is the Mexican presidential election of 2012; below we introduce its, historical background.

Two decades ago, a particular sensibility about media bias has arisen in Mexico, due to the historical and integral role the media played for decades for the authoritarian regime and the State party hegemony (Institutional Revolutionary Party, PRI in Spanish) in terms of quasi-monopoly permits, official advertising, and bribes. Even though the system has slowly broken to allow pluralistic coverage (^{Lozano, 2001}), there still exist mistrust about media performance in politics, not for a lack of valid reasons. In spite of its rapid transition, Mexican media system still can be understood as a "liberal captured" one (^{Guerrero & Marquez, 2014}). The system is embedded in a libertarian model, neutral and objective, based in Constitutional law², but in empirical terms, nationwide and locally, reproduces an authoritarian journalistic culture that extends the editorial subservience of the media to the political actors through clientelist commercialization of their slots (^{Aceves, 2010}; ^{Andrade, 2012}; ^{de León, 2012}; ^{Espino & Mendoza, 2015}; ^{Gonzalez, 2013}; ^{Guerrero & Marquez, 2014}; ^{Ortiz& Gómez, 2013}).

This implies that the norm of objectivity that it upholds is more idealistic than practical (^{Schudson, 2001}), reinforced by schools of journalism (^{Hernández, 2004}), some groups of journalists (^{Marquez, 2010}) and the government and corporate discourses, the latter to emphasize a condition of journalistic autonomy towards skeptical audiences. As these traditions coexist with a critical and politically committed one, though very small, the adherence of the journalists to journalistic norms is rather vague and confusing; therefore, the assessment of media performance becomes difficult and leads to tensions within the stakeholders (audiences, owners, regulators, politicians), which suggests commendable to reconsider the dominant use of neutrality value and adopt an assessment criteria that responds to the actual circumstances and journalistic traditions.

The historical and contemporary problems already discussed gave place, at least from 1988 onwards, to the regular activity of coverage content analysis from academia, civil society, and electoral authority, of a considerable scope and sophistication. In the 2012 election, four NGO´s and fifteen university observatories performed content analysis (^{Buendia, 2013}); nevertheless, these endeavors carry on certain operative understandings, customs, and inertias in the way they measure bias that entails some of the problems discussed here.

Taking these elements as a context of interpretation, we conducted a content analysis that includes these additional practices to those commonly accepted to the observation of media bias: a convenience sample was used, based on the criteria of "base line" and expected coverage (^{Lowry, 2008}; ^{Schiffer, 2006}). This implies to construct a sample of certain facts with a more or less predictable nature that could reveal an editorial stance. Thus, certain events with an agonal component (such as televised debates), politically significant (the opening and closing of each campaign) and bluntly negative to some candidate (fines from the electoral authority to parties, the conflict of the PRI candidate in a University and the rallies that followed, as well as a scandal of illegal fund raising known as "charolazo³)" were selected, based on the assumption that these kind of events are susceptible to party bias. The sample was taken from seven nationwide news outlets (El Universal, Excélsior, La Crónica, La Jornada, La Prensa, Milenio y Reforma), so that 255 news pieces were analyzed (opinion pieces were excluded since they naturally manifest editorial biases, so they are out of the scope of the issues discussed in this paper).

Furthermore, the conventional indicator of "tone" or valence⁴ was complemented with three structural bias indicators: balance, which observes that more than one party or actor is included in the piece so there is an impartial manifestation of world views (^{Carter, Fico, & McCabe, 2002}). Editorialization, that verifies there is an interpretation of the importance and meaning of the events, in a format traditionally devoid of these (^{Niven, 2003}); and the main actor, when the adversary of the protagonist of the piece goes first and gets more visibility; thus functioning as a primary definition (^{Fico et al., 2004}). From ^{Fico et. al. (2004)} we use the strategy of an index⁵ that combines the four indicators and that is compared with that the tone; these practices put forward the difference between many indicators summed up and the use of a single one. In terms of this index, the indicators were transformed as follows:

Where: Y= index result, n=number of indicators, i=indicator and X= normalized value of the indicator

This exploratory proposal has two caveats: it doesn´t give an indicator of visibility inside the news page; and, though it indicates bias, it doesn´t say to whom, something that other auxiliary measures can solve. It is necessary to underscore that the empirical exercise we present is not a case study since the goal is not to delve into the bias practices of 2012 campaign; is merely an example on the advantages of using the sample strategies and indicators already proposed.

Findings

A first relevant finding is about the differences in the extent of the bias for every event in the sample. In an aggregate way, all them are not biased (0.31), but some of them deviate from this mean such as the closing meeting of each campaign (0.43), events of a strong symbolic meaning, the second debate (0.39) with a win-or-lose rhetoric, and the event of the PRI candidate at the Iberoamericana University (0.38) highly contentious as it seems. Next, the coverage of the opening meeting of the campaigns (0.32) and the first presidential debate (0.352); lastly, the anti-Peña rally (0.27) the leftist candidate financial scandal "Charolazo” (0.21), and the fines sanctioned by IFE (0.17), almost without bias. The considerable distance between the latter and the closing events of the campaign, of 26 points in a 100 scale, demonstrate a significant difference in the editorial policies that are exerted to cover different events; particularly those with ample cultural significance or public impact such as presidential debates or campaign closings, and how the specific biases are masked when the data about them is aggregated. In general, by assessing bias for every kind of event, those that were naturally more negative or contested ironically got the more neutral treatment (0.28) in comparison to the presidential debates (0.34) and the opening and closing events of the campaigns (0.37).

Another benefit of the methods proposed is that we can get a complementary, more parsimonious, measure of bias per candidate. A first relevant finding is that candidate Gabriel Quadri, liberal right, got the most biased coverage (0.38) and the leftist Andrés Manuel López Obrador, the least (0.27). Enrique Peña Nieto, center left, (0.30) and Josefina Vázquez Mota, right, (0.32) got similar bias’scores, (0.32) which are moderate. The direction of the bias can be determined by other indicators: under the treatment variable, the coverage was favorable to Quadri (27%) and Peña (10.9%), and less for López Obrador (9.8%) and Vázquez Mota (4.8%). By the indicator of editorialization, neutrality is for López Obrador (90.2%), followed by Quadri (85.7%), Peña (70.5%) y Vázquez Mota (71.4%). The absence of balance, a presentation that benefits some actor by eliminating an opponent in the news piece, is verified for Quadri (78.6%), followed by Vázquez Mota (57.1%), Peña (48.4%) and López Obrador (45.1%). If we observe every indicator, isolated, different interpretations would arise about which party got more bias; instead, the combination of several indicators in a single index makes them look relative and sets a moderate assessment of the extent of the judgment.

Nevertheless, the most meaningful indicator is the candidate’s bias index pertaining the most significant events of the campaign. In the first presidential debate, the less biased coverage was for Peña (0.25), followed by López Obrador (0.31), Vázquez Mota (0.36) and finally Quadri (0.40). In the closing event of the campaign, another highly symbolic situation, the same order is repeated: Peña the actor covered with less biased (0.33), followed by López Obrador (0.43), Quadri (0.50) and Vázquez Mota (0.56). Interestingly, the so called "anti-Peña rally" garnered neutral coverage to the candidates, with an insignificant bias towards Peña (0.17) and López Obrador (0.13), total neutrality in the utterances of Vazquez Mota, and no single piece published about Quadri. At least from the perspective of these three events, Peña got a neutral treatment about a negative event.

Meanwhile, bias ranks per outlet are very variable. There are newspapers with little (Excélsior, 0.11), some (El Universal, 0.21, La Prensa, 0.31 and La Crónica, 0.34) or high bias (Reforma, 0.41, Jornada, 0.43); the first one and La Crónica did not show bias towards any candidate, whereas candidates were benefited in a different way by each newspaper: Quadri with 25% of favorable pieces in El Universal and La Prensa, López Obrador with 50% and 75% of favorable news item in Reforma and Jornada, respectively, and Peña with 40% of positive news in La Prensa.

Lastly, is necessary to adequately assessing media bias from this data. If we proceed by a plain judgment from the official regulations is evident that media do not meet the constitutional standards of objectivity, impartiality, and accuracy; nonetheless, vagueness and lack of operationalization make these principles almost impossible to assess. A second criterion is to evaluate the data in a relative way, by taking some indicators of performance of parties and campaigns that could signal the objective conditions that could lead to unintended bias or a successful adaptation from them to the news making routines. Three aspects can be taken into consideration: differences in campaign funding between parties, this could mean higher professionalism and resources to handle media events; the proportion of seats in the Deputies Chamber at the time of the elections. This could mean a different newsworthiness for each party; and the outcomes of the election, that indicate the "winner" candidates to the journalists and therefore, more newsworthy. In the next table, we present data about each party´s stance in every one of these aspects.

Table 1 Indicators of party salience in campaign 

Source: data from the Federal Institute of Elections (^{IFE, 2012a}).

The table shows an important difference between parties on its political position and resources: the PRI / PVEM (Green Party) alliance is by far the most prominent coalition in Congress (52% of the seats). This gives them a major bargaining power to set the agenda; it took the largest public funding for its campaign, which brings about high public visibility through advertising; and easily won the election. Next, PAN (right) it’s in the second place in regard to the number of seats (29%) but got a similar amount of funding than the leftist parties (PRD, PT, MC), with 26%, although it ends in the third place in the presidential race (26%) way behind the leftist coalition (32%). In terms of political resources and hierarchy, this indicator clearly shows that the PRI / PVEM coalition is in the first place, followed by the leftist PRD / PT / MC and PAN in the last place.

If journalistic routines are sensitive to media events and a significant propaganda environment, both boosted by generous funding; to a hierarchy of parties where some of them are more prominent to the public opinion than others, and therefore, more newsworthy. ; They tend to favor to candidates "on the top of the polls" in a horse race narrative, it could be argued that the lack of neutrality and bias from the media towards the PRI candidate seems related to objective factors of political influence on journalistic routines, rather than a party bias. Lastly, a low consensus journalistic stance about the need for neutrality in political journalism ends up in a scenario of external diversity, where individual bias per outlet neutralizes each other. In conclusion, based on the case examined, and using the research design and assessment criteria proposed, it can be contend that media performance in the coverage of the 2012 campaigns is not heavily biased towards any party. This, argument contradicts public opinion and other scholarly work with the same objectives while using commonly accepted procedures (^{Cantú, 2013}; ^{Martínez, 2013}; ^{Ortiz & Gómez, 2013}). This finding along the procedure to obtaining it demonstrates the value of the methodological proposal presented.

Final words

Citizens’ mistrust in democratic institutions, including news media, generate a public assumption on instrumentality of voters used by political parties as mere information relays of the political elites. Indeed, media bias campaign coverage from campaign to campaign and transgressions on the principles of objectivity and balance that are highly regarded by different actors as well as journalists themselves; nonetheless, as is shown, that relates to far more complex factors that a mere party identification or clientelist force.

This paper illustrates the complexity of party bias and constitutes an invitation to critically think about how convenient are the common methodological techniques that are used to measure it and assess it. The methodological procedure used in this study can, be improved and could get a higher aim if it is applied to non-electoral contexts. However, it represents an argument about the need of a theoretical and methodological discussion, given the importance of the media in political processes, and the need of civil society and the State to review its performance.