Public opinion polls and experts in election news

By employing the theoretical framework of framing, the present paper attempts to examine the Icelandic media’s coverage of the 2013 parliamentary election by paying particular attention to coverage of public opinion polls and the policies of the political parties, i.e. the “horse-race” frame and the issue frame, and to examine media’s reliance on experts for interpretation of election news. Seven online news media, two newspapers, two radio stations and two television channels were monitored for 25 days prior to Election Day, i.e. from April 2 to April 26, 2013, – resulting in 1377 election news stories. The findings show, for example, that 29.8% of all the election news stories had public opinion polls as their primary angle while 12% of the stories were primarily issue-oriented. In addition, the media rely on experts for interpretation of the polls; five of the 10 most interviewed or quoted sources on public opinion surveys were political science experts who were affiliated with universities. Finally, news coverage of polls was generally amplified as media outlets had a tendency to report on public opinion polls that were commissioned by other media.


Introduction
In the summer of 2012, unusual circumstances occurred leading up to the presidential election in Iceland.The incumbent, Ólafur Ragnar Grímsson, was challenged by five other candidates for the office.Public opinion polls soon showed that one of them, Þóra Arnórsdóttir, a well-known television personality, had a realistic chance of becom-

STJÓRNMÁL & STJÓRNSÝSLA
Public opinion polls and experts in election news ing the next president (VB.is 2012).In preparing for a televised debate between the candidates, Stöð 2, a commercial television station, made the decision to invite only Arnórsdóttir and Grímsson to the debate.This decision turned out to be controversial and the invitation was extended to the other four candidates as well (DV.is2012a).However, three of those candidates walked off the stage in front of a live audience at the Harpa Concert Hall and Conference Centre to protest the fact that questions would be directed to the candidates, in alphabetical order, two at a time instead of all six of them simultaneously (Mbl.is 2012).Needless to say, this unexpected turn of events didn't go well with the organizers who commented that the candidates didn't decide the station's programming (DV.is 2012b).
A year later, for the parliamentary election, the media in Iceland were faced with a new dilemma when 11 political parties ran in all six constituencies while two parties ran in two constituencies and two ran in one each (Innanríkisráðuneytið 2013).The Icelandic media had never before had to cover 15 parties before elections and once again, Stöð 2, based its programming decisions on public opinion polls by putting the main focus on the six parties that were likely to surpass the 5% threshold and get their candidates elected to Alþingi, the Icelandic parliament (OSCE 2013).Two of these parties were new, Bright Future and the Pirate Party, while four were returning.On Election Day, the newspaper Fréttablaðið, which belongs to the same media company as Stöð 2, published a large group picture of the leaders of these same six parties and thereby completely ignored the rest of the party chairmen.The not-so subtle message to the newspaper's readers was that the other leaders didn't matter at all in this election.
The Icelandic National Broadcasting Service, RÚV, on the other hand, has a legal obligation to provide political parties, which run in all six constituencies, with an equal opportunity to present their policies (Lög nr. 23/2013).RÚV made an attempt to fulfill its duties by offering all the parties a forum in six issue-oriented television programs and with one-on-one interviews with the parties' leaders, among other things.However, in its final report on the 2013 parliamentary election, the Organization for Security and Cooperation in Europe criticized RÚV for its lack of election news coverage (OSCE 2013).
As discussed above, it appears that the media use public opinion polls as a heuristic device to filter out information, i.e. candidates' messages, which they deem irrelevant to the public because of the seeming lack of candidates' chances of winning.Furthermore, there is some evidence from the 2012 presidential campaign in Iceland that 20% to 40% of the supporters of the "ignored" candidates voted strategically for either Arnórsdóttir or Grímsson (Kristinsson, Indriðason & Valgarðsson 2012).One can only assume that polling results affected the choice of these voters to rally behind a person for whom they would not vote if the situation had been different.
Additionally, findings from a previous study on election coverage in Iceland indicated that the media rely to a great extent on experts, especially political scientists within the academia (Kolbeins 2012).This reliance on experts may be an artifact of the emphasis on public opinion polls as journalists tend to contact experts for comments and interpretations of these polls.

STJÓRNMÁL & STJÓRNSÝSLA
Hence, it is the purpose of the present study to: a) examine the Icelandic media's coverage of the 2013 parliamentary election by paying particular attention to news stories of public opinion polls and stories of the policies of the political parties, i.e. the "horse-race" frame vs. issue frame, and b) examine the prominence of experts.More specifically: R 1 : What is the proportion of news stories on public opinion polls prior to elections?R 2 : What is the proportion of news stories on the policies of the political parties prior to elections?R 3 : What is the proportion of news stories on the opinions of experts and other human sources prior to elections?
To the best of one's knowledge, no study in Iceland has attempted to answer the research questions above -until now.Hence, the present paper throws a new light on election coverage in Iceland and adds to our understanding of journalistic practices in the weeks leading up to parliamentary elections.

Literature review 1.1 Framing of political messages
Theoretically, the context for the current study stems from previous work in political communication which employed framing as the point of departure, and this paper borrows Entman's (1993, 52) much-cited definition of the concept: To frame is to select some aspects of a perceived reality and make them more salient in a communicating text, in such a way as to promote a particular problem definition, causal interpretation, moral evaluation, and/or treatment recommendation for the item described.
However, there appears to be "conceptual confusion" as to what constitutes framing.Scheufele and Iyengar (2014, 7) have argued that "frames have morphed into messages, and the prevalence of emphasis framing in our field threatens to make the broader framing concept redundant as a theory of media effects".They have called for the abandonment of the sociological roots of framing theory and a return to the psychological definitions.Not everyone agrees with this view and it has been argued that it's simply not feasible to adapt the psychological definition of framing because it's not applicable to the real world; that such a narrow definition could only be used in an experimental setting and that it ignores the dynamic aspect that a more sociological definition can provide (Hjarvard 2015).
In part, the conceptual interchangeability of framing and agenda-setting is blamed on Entman's (1991, 7) classic paper where he defines frames "as mentally stored principles for information processing and as characteristics of the news text".Referring to the prominence of news stories, Entman posits that frames can be enlarged and shrunk.Needless to say, most scholars would inevitably argue today that the enlargement and

STJÓRNMÁL & STJÓRNSÝSLA
Public opinion polls and experts in election news shrinking of news stories should simply be considered as agenda-setting (e.g.McCombs & Shaw 1972).More recently, Entman (2007) has urged us to conceive of framing, priming and agenda-setting under the umbrella of bias.
Although the distinction between agenda-setting and framing appears to be blurred, there is some consensus that while agenda-setting tells people what to think about, framing tells them how to think about the issues.Agenda-setting research often explores the relationship between the salience of an issue in the media and the importance of that same issue in people's minds while investigations on framing, on the other hand, reveal how the issue is presented because it may have an effect on how people comprehend it (e.g. de Vreese 2005;McCombs & Shaw 1972;Scheufele & Tewksbury 2007)." Framing […] refers to an active process of creating, selecting, and shaping the frames" (Matthes 2012).
The explication and operationalization of framing as a concept has been a murky business.A good example of this disarray is the fact that coverage of opinion polls, which are generally believed to be an integral part of the "horse-race" frame, have been conceptualized as a distinct frame (Kerbel, Apee & Ross 2000).This certainly lends support to Tewksbury and Scheufele's (2009, 28) claims who said that: communication researchers continue to have an only limited understanding of the more generic sets of frames that can trigger certain underlying interpretive schemas among audiences and therefore lead to various behavioral and cognitive outcomes.

Frame labels
Over the years, it has varied considerably how frames are labeled and used.De Vreese (2012) encourages us to keep in mind that the types of frames can be different.What really matters is that they should be conceptually and operationally clear in every research project.Iyengar (1990) proposed that frames could be episodic or thematic.Thematic frames have information on general trends while episodic frames put personal experience in the spotlight.Frames can also be issue-specific or generic, and most election coverage falls within the second category (e.g. de Vreese 2005).Generic frames can be applied across issues, and among the most common ones are the conflict frame, the issue frame and the thematic frame (Matthes 2009).It's worth noting that the conflict frame can be applied both to news stories that focus on issues and on strategy (Pedersen 2012).Other frames that have often been utilized are the human interest frame, economic consequences frame, morality frame and the responsibility frame (Semetko & Valkenburg 2000).
Many election studies look at news coverage from the perspective of the strategic game frame vs. the issue frame.The strategic game frame covers politics in terms of campaign strategies, public opinion polls and the personalities of the leaders, for instance.Such coverage has a "horse-race" aspect and has a heavy focus on winners and losers.The issue frame, on the other hand, draws attention to the policies of the parties
Although speaking from the sociological perspective of framing, Aalberg, Strömbäck and de Vreese (2011) have called for a consensual approach to the measurements of some of the key concepts of framing in political communication -not least because of its value for comparative research.They propose that the game frame and the strategy frame should be considered separate but equal dimensions.

Factors that influence frames
There is some evidence that country-specific factors, media ownership, type of medium and market share play a role in how much emphasis is placed on the strategic game frame.
Quite a few studies have been conducted comparing election coverage in Sweden to election coverage in other countries, both with similar and different media and political systems.Although the U.S. media tend to be more likely than their Swedish counterparts to cover politics as a strategic game, these differences disappeared when the Swedish public service news was removed from the equation (Dimitrova & Strömbäck 2011).Spanish newspapers have a lesser tendency to apply the "horse-race" frame than Swedish newspapers (Strömbäck & Luengo 2008), which in turn are less likely to do so than U.S. newspapers (Strömbäck & Dimitrova 2006).Some differences even emerged when Swedish media coverage was compared to Norwegian media coverage but the difference in the newspaper coverage of elections as a "horse-race" was insignificant (Strömbäck & Aalberg 2008).The results were mixed when Sweden was compared to Belgium (Strömbäck & van Aelst 2010).
Prior to the EU election in the United Kingdom in 2014, the coverage of the public broadcaster, BBC, was mostly issue-oriented while the privately owned ITV framed its coverage more in terms of the game frame.Moreover, within the game frame, ITV placed a strong focus on the "horse-race" angle of the election (Cushion, Thomas & Ellis 2015).
Despite the claim that competition and commercial factors have lead to the increasing use of the game frame, both in the public service media and the privately owned media, research findings on this subject have not been clear cut.Again, even though it has been shown that public service news is more issue-oriented than news in privately owned media, there hasn't necessarily been an increase in the application of the game frame in news-making (Nord & Strömbäck 2014), and it's not possible to generalize either about the lack of issue-oriented coverage in privately owned media because although there is a tendency for public service stations to have more issue-oriented programming, in some cases the privately owned media did better in this regard, which may be due to regulations (Rafter, Flynn, McMenamin & O'Malley 2014).It should be reiterated that findings of studies that have examined the application of the game frame over time are inconsistent.In Bulgaria, for instance, there was an increase in the use of the game frame from 1990 to 2009 (Dimitrova & Kostadinova 2013).
In general, tabloid newspapers frame their coverage in terms of the game frame while broadsheets are more balanced (Hopmann, Shehata & Strömbäck 2015;Pedersen 2012;Shehata 2014).This has been confirmed in several countries but in Ireland it was also found that the commercial sensitivity of a newspaper in the market decreases issueoriented election coverage (McMenamin, Flynn, O'Malley & Rafter 2012).It appears that in saturated markets with a strong competition for advertising revenue, newspapers are more prone to cover elections as a game than newspapers that operate in markets where there is little or no competition.In addition, the nature of the election campaign itself has an effect.If there is a chance of an upset, there is more emphasis on the election as a contest (Sampert & Petit-Vouriot 2015).

Some consequences of framing
How a particular news story is framed has consequences for the public's understanding of the issue, which is being presented, and how it is perceived.Scheufele and Iyengar (2014) have taken as an example that audiences' reactions to a painting can differ based on whether an art dealer has decided to use an aluminium frame or a gold frame for that same painting.
Strategic news content tends to increase cynicism among young voters who are less politically knowledgeable (Adriaansen, van Praag & de Vreese 2012) and decrease people's internal political efficacy, i.e. their feeling of being able to comprehend politics and to be politically active (Pedersen 2012).Maybe more importantly, being exposed to strategic news decreases people's trust in the media (Hopmann, Shehata & Strömbäck 2015).On the other hand, issue-framing may increase people's political interest and decrease cynicism while game-framing decreases people's institutional trust and their interest in politics (Shehata 2014).However, when the media frame elections as a conflict it may increase voter turnout by mobilizing voters (Schuck, Vliegenthart & de Vreese 2016).
Finally, it should be noted that at least one study has found that people actually seem to prefer the strategic frame because they want to know who is going to win (Iyengar, Norpoth & Hahn 2004).

Effects of public opinion polls
As the present project will show later on, public opinion polls play a pivotal role in election coverage.Yet, they have been heavily criticized for fueling "horse-race" reporting (Patterson 2005) and for providing the audience with repackaged news (Rosenstiel 2005).When covering public opinion polls, journalists often lack the methodological background to recognize inaccuracies in reports from pollsters and they fail to interpret findings correctly when differences are statistically non-significant -not understanding the meaning of margin of error, for example (Pétry & Bastien 2013).Methodological information is also often missing from polling stories (Strömbäck 2009).In its final report, OSCE's Election Assessment Mission on the 2013 parliamentary election in Iceland recommended that the media be more thorough in their reporting of public
Is there a reason to be concerned about the effects of public opinion surveys on voters' behavior?Some evidence has emerged that points towards the band-wagon effect, i.e. that people are likely to vote for those who are ahead in the polls (Ansolabehere & Iyengar 1994;Mehrabian 1998) but what may really matter is the fact that by just tagging polling information on news stories, the media may distract people's attention away from the issues and decrease their ability to acquire information about candidates and parties' policies, even when these are present and salient in the news coverage (Valentino, Beckmann & Buhr 2001;Valentino, Buhr & Beckmann 2001).

Experts in the news
The argument is set forth in this paper that due to the media's emphasis on elections as a game or a "horse-race", they rely on experts for interpretations and predictions of polling results.These experts are often university staff members or affiliated with polling firms.Studies, for instance, in Canada and the U.S. support this claim (e.g.Cross 2010; Freedman & Fico 2004;Freedman, Fico & Durisin 2010).Political scientists have a tendency to refer to polls and tactics in their commentary and hence fortify the "horserace" frame (Brewer & Sigelman 2002).

Election 2013
In 2009, the coalition of the Left-Green Movement and the Social Democratic Alliance became the first left-wing government in Iceland since the country became a republic in 1944, and Jóhanna Sigurðardóttir became the country's first female prime minister (DV. is 2009; Konur og stjórnmál, n.d.).The previous government of the Social Democratic Alliance and the Independence Party had disbanded on January 26, in the wake of several months of unrest and demonstrations after the economic collapse in the fall of 2008 (Forsaetisráðuneytið 2008;Mbl.is 2008;Sigurpálsson 2009).
By 2013, the tide had turned markedly.Shortly before the parliamentary election, public opinion polls were predicting that the Progressive Party was only one man short of gaining a majority in parliament (Mbl.is2013b).The party had taken a firm stand against the Icelandic state guaranteeing payments to the governments of the United Kingdom and the Netherlands because of the Icesave accounts which the now-defunct bank Landsbankinn had offered in these respective countries (RÚV.is2013a).In 2010 and 2011, Icelandic voters had twice in a referendum rejected the negotiations that had taken place between the governments; contracts which members of parliament had passed before the president of Iceland, Ólafur Ragnar Grímsson, used his authority to veto the legislations and refer them to the people (Hagstofa Íslands n.d.a;Taylor 2012).The voters refused to shoulder the debts of a private bank and pay back foreign account holders through their taxes.
In addition, prior to the election in 2013, the Progressive Party promised to decrease

STJÓRNMÁL & STJÓRNSÝSLA
Public opinion polls and experts in election news the capital of people's housing loans by "correcting" them.Due to the inflation in 2008 and 2009, many households had seen their loans sky-rocket, and thereby their monthly payments (Forsaetisráðuneytið 2014;Framsókn n.d.;Skúlason 2014).In January of 2009, the 12-month inflation rate peaked at 18.6%.On average it was 12% in 2009 (Seðlabanki Íslands n.d.).
The events following the crash, and the economic depression, were reflected in the election of 2013.Candidates for 15 parties were running in the election, and among them were such parties as the Households Party and the Iceland Democratic Party which placed emphasis on a new constitution (Mbl.is2013a).
The Independence Party did quite well in the polls too.The party was either leading in the polls or came in second place (Mbl.is2013c).In addition, two new parties were able to grab the media's attention, i.e.Bright Future and the Pirate Party.Other parties that offered a list of candidates in all of the constituencies were the Right-Green People's Party, the Rainbow, and the Dawn.
The Humanist Party, the Rural Party, People's Front of Iceland, and Sturla Jónsson, named after its founder, participated in fewer constituencies.Jónsson was a truck driver who had been outspoken about his financial difficulties during the depression of 2008-2010 and an active demonstrator (Vísir.is2009).
The Progressive Party had nine members in parliament after the election in 2009 but more than doubled that number four years later, going from nine to 19.The Independence Party gained three seats, going from 16 to 19.It had a long-standing relationship with the Progressive Party as they had formed coalition governments from 1995 until 2007, and had worked together in seven other coalition governments from 1947.These two parties formed a coalition government once again in 2013; now with the chairman of the Progressive Party at the helm (Mbl.is2013e).
Four other parties were able to cross the 5% threshold in the election, i.e. the Social Democratic Alliance, the Left-Green Movement, Bright Future and the Pirate Party (Hagstofa Íslands n.d.b)Interestingly, the Pirate Party received 5.1% of the votes but polls had shown that the party might get as much as 7.5%.Thus, public opinion polls had overestimated the party's support (Mbl.is 2013d).
In a nutshell, political development prior to the parliamentary election pointed towards a major government upset and a shift in the public's voting from the left-leaning parties to the middle and to the right.As Sampert and Petit-Vouriot (2015) found in their study of Canadian newspapers, under such circumstances, especially in highcompetition markets, the media are more likely to cover elections as a contest.

Methodology
For the purpose of the present study, all news coverage of 15 online media, print media and newscasts were monitored by the author from April 2 through April 26, 2013 (the day before the parliamentary election), and content analyzed quantitatively.It was thought to be appropriate to start the monitoring on April 2 as it was the Tuesday after Easter Monday.
It should be emphasized that keyword and database search was not applied since all the news stories were either monitored in real time or accessed when they were still easily available and accessible online.This method, i.e. one-person coder and real-time coding, increases the reliability of the study considerably as intercoder reliability is not a problem.Matthes (2009) has criticized framing studies for not reporting on reliability.
A news item was only considered to be an election story if it had a specific reference to the election.For instance, party leaders had to be presented as candidates, or the reader had to be reminded in some way about the upcoming election.Consequently, the researcher's classification of election stories did not always match the media outlets' own groupings of stories, but this methodology resulted in the total of 1377 election stories to be analyzed.
The unit of analysis was always the individual news story.All the items were coded for the date, medium, subject and source, and whether human sources were interviewed or were the topic of discussion.It should be noted that only the first 15 names in a story were coded.This did not affect the results much since only a couple of stories had more than 15 names.In all cases, these were either press releases with lists of candidates or comprehensive coverage on who would lose their seats and who would gain seats in parliament.
Finally, it should be mentioned that the operationalization of the issue frame and the "horse-race" frame were narrower than in many other studies.Election stories were considered to high-light the "horse-race" aspect of the election campaign if public opinion surveys were either the primary angle of the stories or information on how the parties were doing in the polls was added as a secondary angle.To be considered falling under the issue frame, a news story had to provide the voters with information on the parties' policies and platforms, either as a primary angle or a secondary angle.Thus, primary angles were mutually exclusive but secondary angles were not.

"Horse-race" coverage vs. issues
Overall, 29.8% (n=410) of the total of 1377 election stories had polling results as their primary angle.In addition, 7.2% (n=99) had polls as their secondary angle.Twelve percent (n=166) of the stories focused on the platforms of the parties while additional 6% (n=83) had the platforms as a secondary angle.Since coverage of polls and platforms was not mutually exclusive, it was not uncommon for a news reporter to include information on a party's standing in the polls while covering that same party's opening of campaign headquarters and its policies.
The present study did not concern itself with all the stories that were labeled as "other", but it's worth mentioning that these stories covered, for example, the media's own coverage of the election, the turn-out for early voting, the role of the police in the election, interviews with various groups of voters, candidates with criminal records, people who ran into problems with their candidacy, the health care system, and the banks; and last but not least -the weather on Election Day.Looking at the total number of news stories per day, and in particular the number of stories on polling results, one notices the fluctuations (figure 1).As might have been expected, there are generally fewer stories on Sundays than other days of the week, not least because the newspapers have no Sunday versions.Also, there are clear variations in the number of polling stories.This is due to the fact that the media tend to pick up other media's polling stories.For example, on April 5, Fréttablaðið published results from a poll that it conducted in co-operation with Stöð 2. Their sibling-media covered the poll too (Bylgjan and Vísir), but also eight other media outlets.The same thing happened on April 10 when Morgunblaðið reported a survey, which was conducted by the Social Science Research Institute of University of Iceland at its request.Eight other media outlets carried the story, and the website of Morgunblaðið, Mbl.is, of course.Thus, there is a clear indication of an amplifying effect for polling stories.3. 4. 5. 6. 7 . 8. 9 . 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26.

Other
It is particularly interesting to examine figure 2 because it reveals what happened on April 10 and on the following days.At 20:11 on the night of April 10, the website of the business newspaper Viðskiptablaðið, VB.is, reported that 44% of those who supported the Progressive Party would vote for the Independence Party if Hanna
The online media were fast to follow.Eyjan posted the story at 20:37, Mbl.is at 20:59, RÚV.is at 21:48 and Vísir at 22:15.DV.is published the story the next morning, at 7:17, and RÚV in its news broadcasts at 12:20, 18:00 and 19:00.Consequently, on the evening of April 11, when RÚV had planned to interview Benediktsson about his party's policies, the program focused on his decision to step down or stay as the party leader.Thus, a large number of stories on April 12 and April 13, covered the uncertainty of his future and cited the poll by Viðskiptablaðið as a secondary angle.On the morning of April 13, Benediktsson announced his decision to continue as a leader at a campaign meeting in his hometown of Garðabaer.4 . 6. 8 . 10. 12. 14. 16. 18. 20. 22. 24. 26.

Primary angle
As the case of the poll by Viðskiptablaðið so clearly showed, the online media have the advantage of being able to respond quickly when breaking news happen but at the same time, there is a strong need for them to be constantly updating their news cycle.Hence, the online media tend to have more stories than the more traditional media (figure 3).The broadcast media had the fewest election stories; for instance, RÚV at 22:00 only had 27 and Bylgjan radio had 38 altogether while DV.is had 201 and Mbl.is had 169 election stories.
Two interesting comparisons can be made; i.e. between the broadcast media of the privately owned 365 ehf.(Bylgjan radio and Stöð 2) and the newscasts of the Icelandic

Public opinion polls and experts in election news
National Broadcasting Service (RÚV), and between Fréttablaðið, which is also a part of 365 ehf., and Morgunblaðið which is a subscription paper.Four public service newscasts were included in the study and two commercial ones.Hence, 183 stories were analyzed for RÚV and 104 for Bylgjan/Stöð 2.
RÚV and the broadcast media of 365 ehf.placed a similar emphasis on public opinion polls, or 31.1% and 32.7% respectively but the privately owned Stöð 2 was more likely to apply the issue frame than RÚV did.Bylgjan radio only had one story on the issues but Stöð 2 had nine, which meant that Bylgjan/Stöð 2 allocated proportionally 9.6% of the election stories to issues while RÚV had 7.1%.These findings contradict the argument that public service broadcasts are more issue-oriented than purely commercial media.It should be noted that RÚV is not a non-advertising entity.
Looking at Morgunblaðið vs. Fréttablaðið, it appears that the subscription paper is considerably more issue-oriented and slightly less "horse-race" focused than the freeof-charge paper Fréttablaðið which relies more on advertising.Morgunblaðið had overall more election stories than Fréttablaðið (94 versus 68) and more than 18% of its coverage was issue-oriented while 14.7% of the stories in Fréttablaðið had issues as the primary angle.In the case of the newspapers, at least, it could be claimed that less reliance on advertising may increase coverage of issues and decrease, to some degree, the focus on public opinion polls.
Proportionally (figure 4), the different media types tend to carry an equally large share of polling stories while the print media dedicate much more space to stories on the parties' platforms than other media do.Considering the fact that the online media are

STJÓRNMÁL & STJÓRNSÝSLA
not restrained by space, they had the opportunity to cover the issues to a larger extent than other media but did not do that.As stated earlier, there were 15 parties campaigning in the parliamentary election but four of them did not run in all of the constituencies.Election stories that focused on the platforms usually covered several parties, but it was the Progressive Party (B) that was covered the most as a single party (figure 5).The Progressive Party's promise to decrease people's housing loans by repaying them a part of what they had lost due to the inflation in 2008 elicited media attention.Six parties (A, B, D, S, V and Þ) managed to get parliamentary seats and those were also the same parties that got the most media coverage.Four of these parties were returning, i.e. the Progressive Party (B), the Independence Party (D), the Social Democratic Alliance (S) and the Left-Green Movement (V), and two parties were new, i.e.Bright Future (A) and the Pirate Party (Þ).To reiterate, other parties that offered a list of candidates were the Right-Green People's Party (G), the Humanist Party (H), the Households Party (I), the Rainbow (J), Sturla Jónsson (K), the Iceland Democratic Party (L), the Rural Party (M), People's Front of Iceland (R), and the Dawn (T).

STJÓRNMÁL & STJÓRNSÝSLA
Public opinion polls and experts in election news

Secondary angle
Primary angle

The prevalence of experts
Of the total number of election stories that were analyzed, 1039 of them or 75% had a reference to a human being, i.e. an individual's name was mentioned.As many of the names occurred more than once, there was a total of 620 names mentioned.Fifty-six individuals were mentioned at least 10 times and almost half of the names were only mentioned once.Most of the election stories had a single human source and less than half of the total number of stories, or 43% (n=592), had more than two such sources (figure 6).Also, a sole source tended to be interviewed or quoted but if a name had the third position in a story, for example, it was normally mentioned in a discussion between other interviewees and/or the reporter.
Interestingly, 41.4% (n=170) of election stories that had public opinion surveys as their primary angle didn't talk to or mention a human being while only 7.2% (n=12) of the party platforms' stories did so (figure 7).Furthermore, platform stories had more human sources than other types of stories.

STJÓRNMÁL & STJÓRNSÝSLA
Public opinion polls and experts in election news Again, as previously touched upon, Benediktsson, the leader of the Independence Party, was covered extensively by the media during a brief period from April 10 through April 13.This is reflected in figure 8 that plainly shows that he was by far most often mentioned in the media, or 259 times.He was interviewed or quoted 155 times and discussed by others in 104 instances.The vice-chairman of the Independence Party, Kristjánsdóttir, is the anomaly because she was the only one who was discussed more often than directly interviewed.Hence, other people in the media talked about her but the media did not talk to her.One should, however, keep in mind that it is certainly a possibility that she couldn't be reached for a comment ( RÚV.is 2013d).
Moreover, a professor of political science was the only person, among those who were most often mentioned, to whom the media spoke but did not speak about.He was interviewed or directly quoted 57 times, or on average more than twice a day and in 4% of the election stories.It should also be noted that another professor of political science was the tenth most often interviewed person or 30 times.The leader of Sturla Jónsson, Sturla Jónsson, however, got more overall media attention as he was interviewed or quoted 18 times and discussed by other 15 times.Jónsson made the news for various reasons, not least because he was unable to vote for himself as he lived in a different constituency from the one where he was running as a candidate (RÚV.is2013e).Another candidate, Guðmundur Franklín Jónsson, the leader of the Right-Green People's Party, made the headlines when it turned out that he was ineligible to vote in Iceland and couldn't run either because he had resided far too long abroad (RÚV.is2013b).

STJÓRNMÁL & STJÓRNSÝSLA
Looking specifically at interviewees for polling stories (figure 9), it becomes once again evident that the media rely heavily on experts.Four of the ten top interviewees were university staff members of political science, and the fifth expert was a doctoral student of political science.The other five were the leaders of the Independence Party (D), the Progressive Party (B), the Left-Green Movement (V), the Social Democratic Alliance (S) and the Pirate Party (Þ).Four of these parties already had members in parliament but the Pirates were running for the first time and were doing well in the polls.
It should be noted that figure 9 omits all instances where individuals were mentioned in discussions.If these would have been included the picture would have looked a bit different as Hanna Birna Kristjánsdóttir, the vice-chairman of the Independence Party, was the second most talked about person (34 times) in the context of polls but she was only interviewed five times.Moreover, it's particularly interesting that all the experts were interviewed or quoted but never talked about by others.Finally, although the party chairman of the Independence Party (D) was the person who was most often interviewed or discussed by others, the same was not true when it came to the parties' platforms (figure 10).The chairmen of the Left-Green Movement (V), the Progressive Party (B) and the Social Democratic Alliance (S) were interviewed or quoted more often than him.The leader of Bright Future (A) was interviewed quite often too.Also, a candidate for the Social Democratic Alliance, Össur Skarphéðinsson, who was the minister for foreign affairs before the election, was cited or interviewed nine times or equally often as the leader of the Iceland Democratic Party (L).

STJÓRNMÁL & STJÓRNSÝSLA
Public opinion polls and experts in election news

Discussion
To summarize, the present study found ample evidence for the tendency among the Icelandic media to report on elections as a "horse-race" where public opinion polls provide the primary news angle.Although not completely methodologically equivalent, these findings are in line with the results from election studies in Sweden and Norway (Strömbäck & Aalberg 2008) and a Michigan study that showed that journalists' "horserace" experts are often affiliated with political science departments (Freedman & Fico 2004).
Like most research on framing, this project was descriptive in nature and wasn't meant to contribute to framing as a theory (e.g.Matthes 2009).Framing, however, is a dynamic process that has distinct phases, i.e. frame-building, frame-setting and the consequences of framing (e.g. de Vreese 2005).A more in-depth study of framing of election news in Iceland should include journalistic practices, i.e. factors that influence how frames are built within the walls of the editorial rooms, and the effects of these frames on the public.The current study only advanced our knowledge in terms of the content.Furthermore, it focused exclusively on public opinion polls as a part of the "horse-race" and did not include stories on campaign strategies and tactics; nor did it take into account the length of the news stories that were analyzed or their positionwhich are indicators of how much prominence is given to a particular news item.Also, accompanying news pictures or graphs were not analyzed, but the interplay between text and pictures may affect how readers interpret frames.
Interestingly, the data revealed that the privately owned television station, Stöð 2, had proportionally more issue-oriented news stories than the public service newscasts.It may be argued, of course, that RÚV directed most of its election coverage towards general debate programs and one-on-one interviews with the party leaders.Thus, it may have been felt that there was less need to cover the issues of the political parties during the newscasts.These special election programs on television and radio were not included in the present study but their content seeped into the data collection by supplying other media with material for news stories.Stöð 2, of course, also provided its audience with election programs in the debate format.
Why is there so much emphasis on elections as a "horse-race"?It has been posited that strategic game-framing, of which the "horse-race" factor is an important part, is a product of conflict as a news value.Game-framing, and thereby public opinion polls, allows journalists to cover election news objectively, quickly and efficiently, and meet deadlines by using few resources (e.g.Aalberg, Strömbäck & de Vreese 2011;Schuck, Vliegenthart & de Vreese 2016).The lack of human sources for stories on public opinion polls is in accord with this last statement.In addition, these same factors are probably also responsible for the lack of more issue-oriented news stories in online media.Online media have a high turn-over rate of stories, and as there is a constant need to regularly update the websites around the clock, online journalists may have little time to dig into the policies of the political parties.Even though it was not explicitly reported in this paper, a substantial amount of issue-related stories in the online media originated in other media; meaning that journalists, working for either print or broadcast media, were doing a lot of the issue-oriented work for their online colleagues.This gives some support to Rosenstiel's (2005) theory of synthetic journalism where news material from competitors is repackaged by adding something new to it, i.e. journalists practice synthesis instead of original discovery.
But frames have consequences.Game frames increase the public's cynicism and decrease people's trust in the media (Adriaansen, van Praag & de Vreese 2012;Hopmann, Shehata & Strömbäck 2015).Thus, "horse-race" coverage is counterproductive for the media as it's not in their best interest to foster mistrust among their users.Although commercial competition appears to increase the use of the game frame in some cases, it should make more business sense to approach news stories from the issue angle.Again, framing news as conflict can stimulate voting behavior, but conflict and issues are not mutually exclusive (Pedersen 2012).
Framing is conceptually muddy and this study doesn't pretend to make it any clearer.Rather, it used the terminology as scaffolding to base its conclusions on but it's certainly worth contemplating in the future whether the application of framing should simply describe news content, like in this instance and many other such studies in political communication, or whether it is more appropriate to think of framing in terms of providing news consumers with contextual cues that activate certain cognitive schemas -i.e.framing as a theory of psychological media effects.Undoubtedly the debate will continue but presently it is safe to say that as long as the media allocate more time and space to public opinion surveys than to parties' policies and political platforms, and as

STJÓRNMÁL & STJÓRNSÝSLA
Public opinion polls and experts in election news long as the media continue to base their editorial decisions on these surveys, voters may not be acquiring all the necessary information in the election process.

Figure 1 .
Figure 1.The frequency of election stories per day (primary angles)

Figure 2 .
Figure 2. The frequency of primary and secondary polling stories per day

Figure 3 .
Figure 3.The frequency of election stories per media outlet

Figure 4 .
Figure 4. Election stories per type of medium (percentages)

Figure 5 .
Figure 5.The frequency of primary and secondary issue stories per party

Figure 6 .
Figure 6.The frequency of interviewees/subjects of discussion per position in a story

Figure 8 .
Figure 8.The most frequently interviewed/discussed persons

Figure 9 .
Figure 9.The most frequently interviewed persons for polling stories

Figure 10 .
Figure 10.The most frequently interviewed persons for stories on issues . 15.