Newspaper attention and policy activities in Spain

Spain’s newspapers are characterised by strong partisan identities. We demonstrate that the two leading newspapers nonetheless show powerful similarities in the topics of their coverage over time. The media system is strongly related to the policy process and it shows similar levels of skew (attention focuses on just a few topics) and friction (attention lurches rapidly from topic to topic) as others have shown for policy processes more generally. Further, media attention is significantly related to parliamentary activities. Oral questions in parliament track closely with media attention over time. Our assessment is based on a comprehensive database of all front-page stories (over 95,000 stories) in El Paı́s and El Mundo, Spain’s largest daily newspapers, and all 7,446 oral questions from 1996 to 2009. The paper shows that explanations of friction and skew in governmental activities should incorporate media dynamics as well. Political leaders are clearly sensitive to media salience.

the general applicability of a theory developed in the US in a dramatically different media and political context.Spain features a highly politicised media environment and has other features that place it in stark contrast with the US.In their review of media systems, Hallin and Mancini (2004) identify it as a polarised pluralist model, 2 noting that newspaper circulation is among the lowest in the EU, that newspapers are highly politicised, characterised by the low professionalism of the journalists, and that readership for individual papers is divided by partisanship.Spanish newspapers are said to represent distinct political tendencies and to take an advocacy role, mobilising their readers to support different causes.Given that voters (and readers) of the left and right might be mobilised by different issues (social services, retirement and welfare, for example, for left-leaning readers; economic growth and the business environment for the right), we might expect important differences in the topics of newspaper coverage.Jones and Baumgartner (2005) found highly concentrated attention in the low-partisanship US media context.A desire to appeal to rival partisan readerships might create different dynamics in Spanish news coverage, however.
The ideological fragmentation of readers across newspapers that Hallin and Mancini note for Spain has been in place since the mid-1970s, becoming especially intense in the late-1980s with the creation of El Mundo.El Paı ´s was created in 1976 by the media group PRISA, and has always maintained a clear connection with the Socialist Party (PSOE).By contrast, from its creation in 1989, El Mundo became the ''relentless inquisitor'' of the PSOE governments and a point of media support for conservative elites, especially the conservative Popular Party (PP; Castells 2009;Gunther et al. 1999;Reig 2011;Bustamante 2000Bustamante , 2002)).Both El Paı ´s and El Mundo are parts of two of the largest Spanish media groups, PRISA and Unidad Editorial respectively.The PRISA group is the largest media group and also controls one of the main radio stations (la SER), different magazines, and Cuatro, a TV Channel (formerly the pay-per-view TV, Canal 1).Unidad Editorial (controlled by the Italian RCS) is one of the rival media groups that emerged in 2007 after the merger of Grupo Recoletos 3 (owner of the newspapers Expansion, and Marca, the most widely read newspaper specialising in sports), and Unedisa (owner of El Mundo and several magazines like Telva, and a TV network Veo7 that stopped broadcasting in 2012) (Llorens 2010, Jones 2007).
Changing market conditions have not transformed the ideological fragmentation of Spanish newspapers.Still, globalisation of media markets and the consolidation of media companies over the last ten years have fostered the consolidation of a few private medium-sized media groups (Grupo Prisa, Unidad editorial (RCS), Planeta, SA, Vocento, Grupo ZETA, Mediapro and Grupo Godo ´) characterised by an important diversification of activities across different media outlets (newspapers, radio stations, TV), growing internationalisation of activities especially in Latin-America, and the increasing participation of foreign media groups and financial institutions as shareholders (or owners) of national media groups (Jones 2007).
Cross-media ownership is aimed at reducing production costs by achieving economies of scale across multiple media outlets, creating a positive synergy of sharing staff and contents, and increasing revenues with the sale of multimedia advertising packages.This is especially important in the Spanish media system characterised by high levels of fragmentation of the newspaper sector.From the early 1980s to 2010, the number of newspapers has been growing steadily (from 80 in the late 1980s to 110 in the new millennium) although most of them are regional and local newspapers with very small market shares (less than 1 per cent) (de Mateo, Berge ´s and Garnatxe 2010).Five newspapers -El Paı ´s, El Mundo, ABC, El Perı ´odico and La Vanguardia -concentrate most of the readers throughout the period (about 29 per cent in the mid-1980s to 20 per cent in 2010), with some important variations for the case of El Paı ´s and El Mundo.El Paı ´s has always been the leading newspaper, but its market share declined from 18 per cent in 1987 to 12 per cent in 2004 in light of the creation of El Mundo and the growth of other papers.According to Castells (2009, 256) the professional quality of El Mundo and its independence vis a `vis the socialist government provided a platform for the left-wing critics of Felipe Gonza ´lez (in fact, 21 per cent of the readers of El Mundo were voters of the far left in 1993), and made it the second largest newspaper in terms of readership.Since 1996, there has been a clear divide between the readers of both newspapers by party ideology: more than 60 per cent of El Mundo readers are voters of the right, and more than 50 per cent of El Paı ´s readers are voters of the left with few variations across time. 4 Increasing fragmentation of the supply side has not been compensated with a substantial increase in audience.There is a small increment of daily press readers, from about 35 per cent in the late 1990s to more than 42 per cent in 2008, but this is mainly explained by the consolidation of free newspapers -Metro, 20 minutes, Que ´! or ADN -and digital newspapers.The first digital newspapers appeared in the mid-1990s, but it was not until 2004 when this source of information gained significant political relevance (in 2009 the number of readers was 3.7 million versus 1.2 million in 2004).Still, in 2010, 80 per cent of the readers of newspapers prefer the print option, versus 8 per cent who only read digital newspapers.Besides, the consolidation of ICT, and especially the internet, has limited the access of the print media to advertising revenues (in 2008, newspapers only capture 20 per cent of advertising as compared to 30 per cent in the 1980s), a situation that has resulted in financial crisis or even bankruptcy of several newspapers.
In the case of El Pais, changing market conditions lead to an increasing dependency of foreign media corporations and financial institutions.The most important of these alliances is the merger of Telecinco (owned by the Italian group Mediaset) and Cuatro (Grupo Prisa), two of the most important TV channels in Spain, which account for almost 27 per cent of the market share in 2011.By contrast, in 2007 El Mundo (Unidad Editorial) was integrated in full in the Italian media group RCS as a means to overcome financial bankruptcy (Reig 2011).Both cases illustrate the importance of foreign investment in the Spanish media market.After the entry of Spain into the European Economic Union in 1986, most of the big international media groups started to operate in the Spanish market (McChesney 2003).European firms, like RCS (El Mundo), Mediaset (Telecinco), Pierson's group (Recoletos) or Bertelsmann (Antena 3) were among the first to enter the Spanish market, followed in the 1990s by US media groups such as Viacom, Time Warner, or News Corporation, which at present control an important part of the audiovisual sector in Spain (Jones 2007, Berges 2010).Penetration of these foreign media groups in the newspaper sector is not as high as in other media outlets (mainly audiovisuals), with the exception of free newspapers, which are almost monopolised by foreign groups, and the case of El Mundo, which is controlled by the Italian firm RCS (Jones 2007).In all, we can see a clear trend towards increasing concentration and internationalisation of the Spanish media market.The question we pose next is whether these trends generate an increasing convergence on the content of the front pages of El Pais and El Mundo for the last 15 years.

Expectations of divergence or convergence
A ''partisanship hypothesis'' would hold that newspapers would have greater or lesser coverage of individual issues based on which party ''owns'' that topic (see Budge and Keman 1990;Petrocik 1996).This idea of issueownership has been recently applied to the analysis of the media agenda, following a similar argument of existing analysis on party competition (Green-Pedersen and Stubager 2010, Walgrave and van Aelts 2006).Party competition relies not only on drawing attention about different positions on issues, but on drawing attention to issues that have differential benefit to one party, because part of the electorate considers this party better able to handle this issue.According to this idea, in a highly politicised media system we should expect that the media might contribute to the politicisation of some issues, and depoliticise others, drawing attention to issues that are more favourable to the political party with which they are associated.We refer to this as the ''party divergence hypothesis''.
The ''party divergence'' hypothesis is not the only logical possibility, however.There are many reasons why a news outlet, even if associated with a party, may not be able to pick and choose the issues on which it publishes stories.Media outlets are run by journalists, not political leaders, and journalistic norms apply.These imply attention to conflict rather than consensus: the ''horse race'' questions of what political actor is ''winning'' and which is ''losing'' rather than the substance of public policy; and the use of anecdotes and human interest frames, and similar practices (see Graber 2003;Iyengar 1991;Iyengar and McGrady 2007).Bennett (1990) and Gandy (1982) provide further reason to expect minimal differences.For example, Bennett's ''indexing'' theory discusses the patterns of journalistic reporting of ''official'' government actions, as these are seen as inherently newsworthy (see also Sigal 1973).Similarly, Gandy's ''journalistic subsidy'' idea focuses on the tendency for the objects of news coverage to make it easy for journalists to cover them by reducing the costs of stories dramatically.Hamilton (2004) and others make the point that newspapers are bound not only by journalistic norms, but by a business model of capitalism as well.In the search for audience share and profits that go with it, newspapers of all ideological stripes might focus on certain topics (such as ''soft'' news, crime, sensational reporting).Whether from journalistic norms or from adherence to a common business model, all these concepts point to convergence of all media sources on a similar set of topics rather than ideological or partisan preferences driving important differences in the news as seen from different sources.We can refer to this as the ''media convergence hypothesis''.
From a perspective of information theory, or attention scarcity (see Jones and Baumgartner 2005, Baumgartner, Jones and Wilkerson 2011), the range of issues that might be covered in the news is always much greater than the available space allows (Boydstun forthcoming).The hypotheses derived from this perspective also suggest that coverage should be highly concentrated on just a few topics (''concentrated attention hypothesis''), and that patterns of change in attention over time should have the characteristics of friction: heavy replication of the status quo (that is, issues in the news today are highly predicted by what was in the news yesterday), combined with explosive shifts when new topics arise (''friction hypothesis'').Whether these shifts in attention are driven by indexing the actions of government officials, mimicking the actions of other news outlets, or by paying attention to events in the real world, we expect few differences from source to source, but high levels of concentration and friction in both newspapers.
Finally, we test whether differences in issue attention, and levels of concentration and friction in both newspapers are linked to governmental activities.Authors like Bennett (1990Bennett ( , 2004) ) consider newspapers contribute to the politicisation of issues, indexing the activities of public officials, and paying special attention to the issues that are monopolising the political agenda.In this view, the media is seen as a vehicle for governmental officials to criticise each other, but plays no independent contribution to the political debate.The media tends to index the range of voices and viewpoints according to the range of views expressed in mainstream government debate about a given topic, and only occasionally take into account the points of view different from the mainstream official view of the government and political parties.Accordingly, we should expect the media agenda to be partly driven by the political agenda.Specifically, when the number of oral questions in Parliament on a topic increases we expect to see an increase in media attention in the following week (''indexing hypothesis'').
A media salience model would stress that the media plays a more important role in leading the political agenda.Authors like Castells (2009) or McChesney (2003) argue that globalisation and increasing market competition plays a key role in facilitating the close interaction between the media and politicians.
''News outlets need political actors to deliver sensationalist stories that attract audiences as much as they need political decision-makers to relax regulation and conglomeratisation laws.At the same time, politicians need media organisations to deliver their messages to the public in a way that activates the median voter.''(Arsenault and Castells 2008) This interdependence is shaped by market forces and reinforced by the structural configuration of news-making and journalistic conventions.According to this, we should expect the media to have some autonomy in setting its agenda, prioritising some issues and not others in order to maximise readership and enlarge audiences.Further, political leaders may follow, not lead, the media.When stories gain great media salience, political leaders may increase their activities in those areas in order to persuade voters that they share these concerns or perhaps simply to gain more public exposure (''media agenda-setting hypothesis'').

A comprehensive analysis
In order to test the various hypotheses we lay out above, and more generally to understand the nature of news coverage of politics in Spain, we have already created a comprehensive and far-reaching dataset.It includes every article from the front pages of two of the most relevant Spanish newspapers -El Paı ´s, and El Mundo -from 1996 to 2009.The development of these databases follows the methodology of the Comparative Agendas Project (see www.comparativeagendas.info), a methodology that has already been used by this research team for the creation of databases about the political agenda and public opinion in Spain; here we focus only on the media and oral questions, however.As is typical in the agendas project, we have created a comprehensive database, not a sample.Each story, picture and illustration that appeared on the front page in either of the two papers during the 14-year time span of our study has been included, approximately 100,000 stories in all.
For each story, we have recorded the title, date of publication, size of headline, story and photo, information to identify the most important actors or institutions mentioned (ranging from local governments to national and international actors), political parties mentioned and the topic.The topic of the story is classified according to the 23 major topics and 247 subtopics of the Spanish Policy Agendas project (www.ub.edu/ spanishpolicyagendas).We have also created other variables to analyse whether stories refer to elections (national, regional, local or European) morality issues (such as abortion, same-sex marriage) or political scandals (by type of scandal and political party involved).As a whole, for each article we have gathered information about 37 different indicators, which allow for systematic coding of media agenda.Each story has been coded by two different coders with a reliability of 93.25 per cent at the topic level and 90.35 per cent at the subtopic level.Appendix 1 provides more information about the Spanish Agendas Project.
To test the interrelation between the media and political agendas, we focus on the particular case of oral questions posed to the Presidente del Gobierno and other members of the executive in the Spanish parliament in plenary meetings.This is a good indicator to analyse the correspondence between the media and the political agenda, which is mainly short term.Oral questions are one of the main instruments available to members of parliament (MPs) for controlling what the executive is doing, providing crucial information about the issues that are more politicised across the legislature (Vliegenthart and Walgrave 2011).
The introduction of an oral question is quite an open process in which a single deputy or a parliamentary group can get involved.MPs use oral questions as devices to publicise good (or bad) news, to give visibility to issues they think are important or (if they support the governing party) to delay reactions to bad news, which may have high electoral cost.The capacity of a parliamentary group to ask one or more oral questions in a control session depends on the number of seats, which in practical terms means a clear division between the two largest political parties (PSOE and PP account for more than 80 per cent of the seats) and the rest of political parties. 5 To ensure that we have our time order correctly specified, it is important to know the timing of oral questions in the Spanish parliament.Questions must be submitted by Monday morning for the weekly oral question session, which occurs on Wednesdays.We can therefore test whether the media are setting the political agenda, or rather indexing political activities by comparing newspaper coverage in the previous week to the oral questions posed.If the media lead the questions (that is, a positive relation between articles in the previous week with questions in the current week), we confirm the media agenda-setting hypothesis.If the relationship is statistically contemporaneous (questions on Wednesday are related to articles in that same week), then we disconfirm the agenda-setting hypothesis and confirm the indexing idea.Of course, there may also be no relation.

Results
Spain indeed has a fragmented and ideologically distinct set of newspapers, and the two papers on which we concentrate have clear partisan identities.In this section, we review a range of empirical results that test the hypotheses laid out in the first section.

Topics of coverage
Table 1 shows the distribution of front-page stories by topic.We define 23 distinct topics of attention (and further subdivide these into some 247 more detailed subtopics as discussed above).The table shows that the main topics of attention are the same in both papers: justice and crime (21 per cent of the stories in El Mundo; 16 per cent in El Paı ´s), government operations (20 per cent; 17 per cent), culture and arts (9 per cent; 10 per cent), sports (7 per cent; 6 per cent), defence (6 per cent; 7 per cent), and rights (6 per cent; 7 per cent).Similarly, those topics, such as foreign trade, agriculture, and public lands and water, that receive very little attention in one paper also receive little in the other.Figure 1 shows the data from Table 1 in graphical form, making clear the high correlation between amounts of coverage across the 23 topics; the correlation in fact is 0.98.
A review of these data suggests that while there are some differences between the papers (more crime coverage in El Mundo, for example, more defence and international stories in El Paı ´s), readers of the two papers are getting generally a very similar mix of stories.This suggests that the partisan difference hypothesis may be incorrect.Rather, the evidence is strongly supportive of convergence rather than partisan difference in newspaper coverage. 6 These differences make clear that indexing, if it occurs, is selective.Much of the routine oversight of government that goes on in parliamentary question time is simply not newsworthy.Media attention is concentrated on those topics that generate higher public interest.Parliamentary attention is more broadly spread across all the topics of public affairs.

The dynamics of coverage
Not only are the topics of coverage highly correlated, but they appear to follow the same patterns over time.Figure 2 shows the number of stories per month on six select issues.The upper-left graph in Figure 2 shows the number of stories per month on the state of the economy.The economy typically has about ten stories a month on the front page of either newspaper, but there are noticeable periods of heightened and lower attention; these are highly similar in the two papers.Agriculture, in the upper-right graph, shows lower attention overall, but nearly identical peaks of attention in both papers during those rare periods when it hits the front pages.Across all the examples in Figure 2 (and similarly in the cases not shown, for lack of space), the number of stories in one paper is a very strong predictor of the number of stories in the other.The papers may be mimicking each other or they may both be responding to the same cues in the environment, but there is no support for the idea that partisan differences in ideological orientation lead one paper to report on topic X while the other paper is focusing on topic Y. Again, it is clear that convergence, not divergence, is the rule.The similarities do not stop at only the choice of topics.The two papers are virtually identical with regards to coverage of policy-relevant versus non-policy topics.Figure 3 shows that the papers are almost identical in the proportion of arts, culture, sports and human interest stories that they publish on the front page, as compared to the ''hard news'' of government and public-policy relevant information.(We also see no trends over time in the percentage of ''soft news'' though we do see a cycle where there is much greater attention to these topics each year during the vacation months of July and August.) In sum, a review of the topics of attention and of the percentage of soft news shows that there are very few differences between the two papers.The similarities continue when we look at the concentration of attention on just a few topics, and on patterns of how attention shifts from topic to topic over time.

Highly concentrated attention
While there may be any number of issues subject to public concern or to official government actions, only a few reach the threshold of newsworthiness to hit the front pages.Figure 4 shows that both newspapers concentrate their attention on an extremely small number of topics.
Looking at the right side of Figure 4, it shows that over 10 per cent of all the stories in El Mundo were on a single topic and that only a few topics had over 2 per cent of coverage, with the rest of the topic areas defined in the Spanish Agendas Project receiving miniscule proportions of the total coverage, or no coverage at all.The left panel in Figure 4 shows an almost identical pattern for El Paı ´s. Figure 5 and Table 2 show exactly what those topics of high attention are, for the two papers.
Figure 5 lists the topics receiving more than 1 per cent of the total coverage.As is clear, the topics are similar in the two papers.The topic numbers in the legend of the figure are reproduced in Table 2, which gives a short description of the topics that dominate attention in the two papers.
Table 2 lists the top 20 topics of attention and makes clear that terrorism, corruption, war, political parties and campaigns dominate the front pages, as do certain broad categories such as sports, culture and performing arts.The cumulative percentages listed make clear that just 20 topics (out of 247 available in the Spanish Agendas Project topic system)   Attention in both papers is highly concentrated on just a few topic areas.
constitute 50 per cent of the agenda for El Paı ´s, and 56 per cent for El Mundo.Further, as is clear by perusing the table, the individual topics may vary slightly in order, but they are highly similar across the two newspapers.So we see very strong support for the concentration of attention hypothesis, and only minor differences between the two papers both in the topics of highest attention and in the degree of skew in coverage.

Stability and explosive shifts in attention
Attention naturally changes from topic to topic as the news evolves.We saw in Figure 2, for example, that attention to any particular topic can be highly unstable, moving up and down in a particular month as events, other news, and the actions of newsworthy actors in the political system affect the newspapers.Naturally, some topics typically gain more attention than others.Looking at how attention shifts over time, some months might be expected to produce slightly more Economics news than others, for example, or less Foreign Trade news than the month before.If there were no institutionalised routines, and if the events in society were evolving according to a large number of different and unrelated factors, then the Central Limit Theorem would guarantee that the distribution of changes in attention would be approximately Normal, producing a bell-shaped curve (see Jones and Baumgartner 2005 for a more detailed explanation of these dynamics).Figure 6 shows the relevant distribution of month-to-month changes in attention to the 23 major topics in the Spanish Agendas Project.
The figure shows very similar patterns in both newspapers suggesting a high degree of friction in the process of news-making.That is, changes are highly clustered around zero, meaning that the number of stories in a given The distributions show that, from one month to the next, the number of stories on a given topic is typically very similar to what it was in the previous month (for example, the monthly change value is close to zero).However, the distributions also show extremely great change in a relatively large number of cases.These are when the papers publish a great number of articles on a topic that they had relatively ignored in the previous month.K and LK scores show that the distributions are far from a normal distribution, which would occur if the monthly changes were random.
month is likely to be very close to the number in the previous month (in the figure, we eliminate all observations where there are no stories, to avoid an over-inflation of zeros based on no coverage at all).But when new topics emerge on the front pages, shifts in attention can be highly explosive.The distribution is simultaneously peaked at zero and with a ''fat tail'' going out to the right, evidence of a process with great friction: little change most of the time, but explosive rather than moderate change when it occurs.

Links between media coverage and parliamentary questions
Because Figure 1 showed that the topics of attention were correlated almost perfectly between the two newspapers, we take the average percentage across the two newspapers for each topic to make an estimate of the ''newspaper agenda'' and we compare that with the ''oral question agenda''.Before looking at the dynamics over time, Figure 7 shows the difference in topics between newspapers and oral questions.
The left-hand figure shows the percentage of all front-page stories and the percentage of all oral questions focused on each of the 23 major topics of the Spanish agendas project.In the right-hand figure, we subtract the percentage in the oral questions from the percentage in the media.Thus, for example, justice (such as stories about crime) represents about 18 per cent of the newspaper stories but only 9 per cent of the oral questions and therefore appears with a value of 19 in the right-hand figure.This presentation of the data makes clear that four topics are more common in the media than in the oral questions: justice, rights, defence and government (such as elections).By contrast, members of parliament spend considerable time focusing on the routine matters of public policy such as transport, health, energy and foreign trade, all of which are less interesting to the media, or at least are relatively rarely seen in the front pages.These topics are noted by the negative values in the right side of Figure 7, and they represent the bulk of the cases.Media coverage is more highly focused on a small number of issues, with justice and government issues each receiving around 20 per cent of the total attention whereas oral questions are more evenly spread, with no single topic receiving as much as 11 per cent of the total number of questions.
The characteristic patterns of skewed attention to a few topics, common attention to the same topics across newspapers, and explosive rather than gradual change from topic to topic could potentially be due to newspapers taking their lead from government officials.Others have shown that government attention shares these characteristics (see Baumgartner, Jones and Wilkerson 2011).In this section, we explore the links between media coverage and parliamentary questions.Earlier, we reviewed two important perspectives on the possible linkages we might observe.
According to indexing theory (Bennett 1990), we should expect that the media: ''tend to index the range of voices and viewpoints in both news and editorials according to the range of views expressed in mainstream government debate about a given topic''.(Bennett 1990, 106) The media is a vehicle for governmental officials to criticise each other, which only plays an independent contribution to the political debate when there is a high level of controversy about a specific issue (Entman 2004, 4).The media should give information about what is going on in politics, what the issues discussed in the political arena are and how they change over time.Beat reporting, the fact that government officials are considered to be newsworthy and other factors of the news gathering process suggest that media coverage might simply reflect official activities.To test the indexing theory, we simply compare the topics of parliamentary activity (oral questions, in the present analysis) with media stories in the following week.If indexing is going on, we will see the media printing stories on the same topics that the deputies are asking questions about.
A rival perspective is the media agenda-setting hypothesis.Here, the deputies follow the lead of the media, seeking to gain advantage from the fact that certain issues are highly salient in the media.We test for this by assessing the link between questions on a given topic with the number of newspaper stories on that same topic in the previous week.
Table 3 shows the results of a series of simple regressions demonstrating the statistical linkage between media coverage and parliamentary questions.For each policy domain, we simply ask whether the percentage of questions posed on that topic can be predicted by the percentage of front-page stories on that same topic in the preceding week, or if the newspaper agenda, by contrast, can be predicted by the oral questions.
Table 3 provides a direct test of the indexing and the media agendasetting hypotheses, and it is clear that agenda-setting is much more prominent than indexing, though both occur.Parliamentary questions are posed in the Spanish parliament on Wednesdays, but the questions must be submitted in writing on Monday mornings.Therefore, we test the agenda-setting hypotheses with a regression that predicts the percentage of total questions by a constant (results not shown) and the percentage of media stories in the previous week.Significant (p , .05)results are shown in bold; 16 of 19 relations are significant, and the three that are not significant are all cases where the level of media attention is typically almost zero.Results are particularly strong for the issues of environment, economy, defence and energy.
With regards to the indexing hypotheses, the results are much weaker, with significant relationships only in nine of the 19 cases.The strongest The two columns on the right show the mean percentage of questions or media stories per week on each topic.The models are identical in each case.For the media agenda-setting model, the percentage of all oral questions in the topic area is predicted by a constant (not shown) and the percentage of all front-page newspaper stories in the two newspapers on that same topic in the previous week.Sixteen of 19 coefficients (shown in bold) are significant at ,.05.Note that for the three cases with insignificant results (social affairs, foreign trade and public lands), there is virtually no media coverage.For the media indexing model, the percentage of all front-page newspaper stories in the two newspapers in the topic area is predicted by a constant (not shown) and the percentage of oral questions on that same topic in that same week.Nine of 19 coefficients (shown in bold) are significant at ,.05.Data cover the period of 1996 to 2009.Parliament is not in session in all weeks, explaining why the N (361) is less than the total number of weeks during the period.Eighteen per cent of media stories are on topics unrelated to public policy (such as sports, movies), and those are excluded here.Among parliamentary questions, 99.7 per cent are included here.
relations are in the cases of environment, defence, economy, energy, health and agriculture.This suggests that in these cases there is evidence that media coverage is gauged on the parliamentary salience of the issue.It is, of course, not surprising that some issues would show evidence for both indexing and agenda-setting, as both can occur and surges in attention may last more than one week in any case.However, the data in Table 3 do make clear that agenda-setting is a much more powerful effect.The coefficients themselves can be interpreted to make this imbalance even clearer.A 1 per cent increase in media stories on the economy, for example, is expected to lead to a 1.42 per cent increase in questions, but the reciprocal change is expected to lead only to a 0.06 per cent increase in media stories.So the media have much greater effect on the deputies than the deputies do on the media.Figure 8 illustrates the patterns between media and parliamentary activity by showing the percentage of each across six policy domains.The figures show monthly data rather than the weekly data in the regression analysis, in order to smooth out temporary fluctuations.
The upper-left graph shows the case of environment, which showed the highest relationship in Table 3.While it is certain that the ecological disaster of the grounding of oil tanker Prestige would have elicited substantial parliamentary investigation in any case, the extensive media coverage of the event certainly contributed to it.Similarly, we see in the cases of energy, defence and macroeconomics that questions are substantially correlated with media attention.Of course, the table also makes clear that some policy domains are relatively unaffected by these dynamics and at the bottom of Figure 8 we illustrate this with the cases of health and social affairs.
When members of parliament ask oral questions to the ministers, they tend to follow what is in the news.Our data show that media attention drives parliamentary question time.But the analysis has also shown that the two main Spanish newspapers vary only slightly from one another in terms of their topics of attention.The convergence of media coverage on a few topics in any given month of the political calendar, whether driven by journalistic norms, by real-world events or by commercial considerations, means that the two leading newspapers in Spain are not systematically cherry-picking only those topics of attention that make their partisan allies look good.Rather, they consistently follow very similar agendas.

Conclusions
The Spanish media system is highly partisan.Readership and editorial preferences are substantially divided by party preference.In this environment, it is easy to expect that the papers could contribute substantially to a system where supporters of different parties experience politics in substantially different ways.Journalistic norms of what is newsworthy, on the other hand, require that major events receive treatment.We find, based on a large-scale assessment of the content of all stories appearing on the front pages of El Paı ´s and El Mundo, that the papers differ little in many important aspects: they cover the same topic areas, they cover similar percentages of ''soft news,'' they focus attention equally on certain topics to the detriment of others and they show substantial ''friction'' in how they shift attention from topic to topic over time.In all these ways, the papers are highly similar.
Our results also illustrate that these features have been consistent for the last 15 years, without much variation across time.Increasing market fragmentation, internationalisation of activities, and the penetration of foreign investment in the Spanish media market have not generated major changes in the way the two leading Spanish newspapers prioritise across issues.Contrary to some expectations, globalisation and increasing market competition have not led to changes on the media agenda towards more stories without political content (such as sports and entertainment).
We also find a strong correspondence between the media and the political agenda as measured by oral questions.There are powerful connections between media coverage and parliamentary behaviour, even if the deputies regularly work in areas that are not very newsworthy.When media attention spikes, as it often does, parliamentary attention follows.Thus, the media appear to be important contributors to the fact that government attention cannot be predicted solely by looking at the severity of various social problems.Rather, it lurches from topic to topic based at least in part on what is on the front pages.By the same token, increased attention in parliament is also followed by increasing attention by the media, although the direction of this connection is much weaker.That is, we find much more media agenda-setting than we do indexing.
In sum, the Spanish media system exhibits characteristics that fit into a general model of attention scarcity.The highly partisan structure of the media may have important impacts on the substance of its coverage.However, it has no impact whatsoever on the common finding of concentrated attention and powerful threshold or friction effects.These appear to be ubiquitous characteristics of political and media institutions everywhere.The tendency of media sources to concentrate attention on just a few topics has powerful effects on government, exacerbating already strong tendencies of political leaders to lavish attention on just a few areas of public policy while ignoring completely many others.NOTES

FIGURE 2 .
FIGURE 2. Monthly stories on selected topics in two Spanish newspapers 1996-2009The figures show the number of stories per month on selected topics in El Paı ´s and El Mundo from 1996 to 2009.The number of stories on each topic is typically highly correlated between the two newspapers.

FIGURE 3 .
FIGURE 3. Policy and non-policy coverage

FIGURE 4 .
FIGURE 4. The distribution of stories by subtopic In both newspapers, coverage is highly concentrated on just a few subtopics.

FIGURE 5 .
FIGURE 5.The topics with greatest attention, by newspaper

FIGURE 6 .
FIGURE 6. Patterns of change in attention in two Spanish newspapers 1996 to 2009

FIGURE 7 .
FIGURE 7. Newspapers and parliamentary agendas

FIGURE 8 .
FIGURE 8. Media coverage and parliamentary questions in selected areas

TABLE 1 .
News coverage in El Paı ´s and El Mundo, by topic, 1996 to 2009 Source for all tables and figures: Spanish Agendas Project.FIGURE 1. Topics of coverage in two Spanish newspapersSee Table1for exact numbers and percentages.Bonafont and Baumgartner

TABLE 2 .
The Concentration of Attention in a Small Number of Topics