VOX POPULI OR VOX MEDIA?
OPINION POLLS AND THE SWEDISH MEDIA, 1998-2006 JESPER STRÖMBÄCK
Abstract
While the histories of opinion polling and the news media have been closely intertwined ever since the invention of polling, the question as to whether the media's reporting on opinion polls should be considered as detrimental or beneficial from a democratic perspective is still open and contested. The purpose of this paper is thus to investigate the publication of opinion polls in the Swedish media during the last three election campaigns, with a focus on how the media used opinion polls and whether or not the media, at the end of the day, mainly used opinion polls to give voice to the people - or to the media and the journalists themselves. Among other things, the results suggest that more often than not, polls serve as vox media
rather than vox populi.
Jesper Strömbäck is Professor in Media and Communication in Mid Sweden University, Sundsvall; e-mail: Jesper.Stromback@miun.se.
O
m m
d
ro o
o o rN
£
i 0 £
II) G C
(6
Ln Ln
Introduction
The histories of opinion polling and the media have been closely intertwined ever since the invention of polling. Whether the media's reporting on opinion polls should be considered as detrimental or beneficial from a democratic point of view is, however, open to debate (Glasser and Salmon 1995). As always there are both critics and defenders. With regard to the defenders, they argue that the publication of opinion polls send a symbolic message that the opinions of "everyone" matters (Lavrakas and Traugott 2000), that it empowers the media to serve as independent watchdogs of those in power and to speak on behalf of the public (Gollin 1980; Ladd 1980), and that it raises the public's interest in political campaigns (Iyengar et al. 2004). With regard to the critics, some are critical of opinion polls per se or argue that they mainly serve the interests of political and media elites (Bourdieu 1979), whereas others focus their criticism on the frequency to which polls are published (Patterson 1993), how they are reported (Bishop 2005; Bogart 2000), the use of causal explanations in interpreting polls (Bauman and Lavrakas 2000) or how this kind of reporting spurs the framing of politics as a horse race (Cappella and Jamieson 1997; Patterson 2005). On a theoretical level, the relationship between public opinion as such and as measured by opinion polls has also been questioned (Salmon and Glasser 1995; Herbst 1995; 1998; Lewis 1999; Splichal 1997; 2008). One of the consequences of the rise of opinion polling is that public opinion as measured by polls has become equated with public opinion per se, and that other purveyors and representations of public opinion - such as parliaments, political parties, interests groups and partisan media - have lost legitimacy as or even disappeared from our understanding of public opinion. It has shifted our understanding of public opinion, and hence the role of public opinion in democratic societies.
The focus of this article is however not the larger theoretical debate about public opinion and opinion polling, but rather the media coverage of opinion polls. For most people, reliant on the media for information about matters beyond their everyday experiences, the media's coverage of opinion polls help shape their y understanding of public opinion, and for elite groups, both the own use and the media's coverage of opinion polls are crucial in their understanding of people's opinions and attitudes (Herbst 1998).
More specifically, the purpose of this paper is to investigate the publication of opinion polls in the Swedish news media during the 1998, 2002 and 2006 election campaigns, with a focus on how the media used opinion polls in their election news coverage and whether or not the media, at the end of the day, mainly used £ opinion polls in order to give voice to the people - or to the media and the journalists themselves.
I
(I) 0 C
A Matter of Frequency
Once described as the "pulse of democracy" (Gallup and Rae 1940), the usage of opinion polls by the media has attracted critics as well as defenders. Notwithstanding the fact that "Public opinion continues to be one of the fuzziest terms in the social sciences" (Donsbach and Tragugott 2008) and the debate regarding what opinion polls in fact measure and represent (Salmon and Glasser 1995; Splichal 1997; Lewis 1999), one line of criticism is that the media focus too much on opinion ^ polls. Patterson (2005, 722) for example, argues that the use of opinion polls has
extended "beyond reason," while Weimann (1990) writes about an "obsession to forecast." Although there is no standard for evaluating when the media focus too much on opinion polls, evidence from the United States (Traugott 2005) as well as ^ Germany (Brettschneider 1997) and other countries (Brettschneider 2008; Weimann 1990) suggests that there has been a major increase over time in the media's coverage of opinion polls. In the U.S. case, Traugott (2005, 644) has shown an increase of about 900 percent between 1984 and 2000, while in the German case, Brettschneider (2008, 482) has shown that the number of poll reports has increased from 65 in 1980 to 651 in 2002.
In the Swedish case, evidence is more mixed. According to Petersson and Holmberg (1998, 116-117; Holmberg 2008), the frequency of articles reporting results from opinion polls peaked in the mid 1980s, after which it declined and then rebounded. Considering the time span of this particular study and trends in the media's focus on opinion polls in other countries, the first hypothesis is that there has been an increase in the number of news stories in the Swedish media that report results from opinion polls: The number of news stories in which opinion polls were reported increased over the election campaigns 1998-2006 (H1).
A Matter of Object and Framing
The frequency of news stories focusing on reporting on opinion polls notwithstanding, of equal interest is the main object of the polls and how the publishing of polls shape the media's framing of politics. In theory, the media's coverage of opinion polls can be perceived as a means to provide voters with important information on how people in general think about various matters, ranging from policy proposals to how they intend to vote on Election Day. Polls can be used to inform the people about itself on the most important issues of the day.
However, most research suggests that polls on people's voting intention dominates the media's poll coverage, at least during election campaigns, and that this contributes to the framing of politics as a horse race or a game rather than in terms of issues and policies (Broh 1980; Patterson 1993; Cappella and Jamieson 1997). As suggested by Rosenstiel (2005, 710), "More stories about the daily horse race shift the focus of the race. More horse race polls, in short, translate into more horse race coverage." Patterson (2005, 718) similarly observes that "the use of polls spurs horse-race reporting - the tendency to treat elections as if they were sporting events where the paramount goal is to get across the finish line in the first place," while Atkin and Gaudino (1984, 124) conclude that "Instead of covering the candidates' qualifications, philosophies, or issue positions, polls have encouraged journalists to treat campaigns as horse races, with a focus on the candidates' popularity, momentum, and size of the lead" (see also Brettschneider 1997; 2008; Farnsworth and Lichter 2007; Petersson et al. 2006). From this perspective, publishing opinion polls distracts the media from focusing on the most important issues of the day and on information that people might need when forming their opinions and voting preferences. The framing of politics as a strategic game has also attracted criticism based on evidence suggesting that such framing increases people's political distrust and cynicism (Cappella and Jamieson 1997; de Vreese and Elenbaas 2008).
Thus, while it certainly is legitimate to publish horse race polls and frame politics as a game, an excessive focus on the horse race and horse race polls might
overshadow the issues at stake in an election, and thus contribute negatively both to people's trust in political actors and institutions (Cappella and Jamieson 1997) and their opportunities to use the media to gain information and cast enlightened votes (Rosenstiel 2005).
Also important in this context is that standard horse race polls have limits when it comes to informing the public, the media and political actors about people's thoughts, priorities, issue positions and values. Such polls "heighten journalists' attention to the candidates, rather than to the voters themselves," as noted by Patterson (2005, 720). Consequently, more often than not horse race polls might fail to give those whose opinions are polled a clear voice in political communication processes.
All polls are not equal, however, and the discussion above raises two questions. First, what was the object of the polls that were published in the Swedish media during the 1998, 2002 and 2006 election campaigns? Were the polls about party or candidate preferences, issues and issue positions, or other objects? Second, is there a correlation between the media's publication of opinion polls and the framing of politics as a game?
Based on prior research it can be expected that most published polls focused on the horse race and that there is indeed a correlation between the publication of opinion polls and the framing of politics as a game (Atkin and Gaudino 1984; Patterson 1993; 2005; Rosenstiel 2005; Farnsworth and Lichter 2007; Brettschneider 2008). Thus, the next two hypotheses are: (2) Most of the published opinion polls during the election campaigns 1998-2006 were about people's party or party leader preferences (H2); (3) There will be a positive correlation between the publication of opinion polls and the framing of politics as a game (H3).
Matter of Interpretation
The notion that opinion polls contribute to the framing of politics as a game and heighten journalists' attention to the candidates rather than the voters suggests that the media's interest in polls is driven less by a concern with voters and y their opinions, and more by the media's own needs. Polling can be an important newsgathering tool (Ismach 1984), and by commissioning polls, the media secure for themselves a triple role: they commission the polls; they report them, and they interpret the results (Petersson et al. 2006).
Thus, by commissioning polls the media not only make their own news (von Hoffman 1980) - they also grant themselves the privilege to define and frame the actors, events or issues being polled. This privilege can be used to gain independence from and question political power holders (Gollin 1980; Ladd 1980; Lang and Lang 1980). As noted by Lavrakas and Traugott (2000, 10), polls can aid democratic processes by "Empowering the media to serve as an independent watchdog on politicians and resisting other would-be spokespersons for the public or for so-called election mandates." They also note (2000, 4) that one of the main reasons for why journalists are attracted to polls is that these allow journalists "a quasi-objective, proactive role in the newsmaking process."
Thus, there appears to be a linkage between the publication of opinion polls and an interpretive journalistic style (Ismach 1984; Patterson 1993). Rosenstiel (2005, 706) consequently notes that: "more and more journalism involves synthesising that competitive material into one's own account and then adding something new
0
fi ¥
I
C >
or special to it or trying to account for all that information into one interpretive or analytical frame." The framing of politics as a game is one such frame, but if polls C^ are attractive to journalists because they allow them a quasi-objective, proactive role, then the linkage between poll reporting and an interpretive journalistic style should extend beyond this particular framing of politics. Considering this and evidence suggesting that an interpretive journalistic style has become more common in many countries around the world (Kovach and Rosenstiel 1999; Patterson 1993; Strömbäck and Kaid 2008) the fourth hypothesis is: There is a positive correlation between the publication of opinion polls and an interpretive journalistic style (H4).
A Matter of Methodological Information
Regardless of the object of published opinion polls, the framing of politics and the journalistic style employed in news stories on polls, from an informational perspective it is important that the publication of opinion polls is accompanied by enough methodological information for people to be able to critically assess the polls (Welch 2002). Although it can be questioned whether most news consumers understand and make use of methodological information on published polls (Traugott 2004; Wichmann 2008), it is generally accepted that the validity and value of opinion polls is dependent upon "the way in which the findings are presented and the uses to which they are put," as stated in the ESOMAR/WAPOR International Code of Practice for the Publication of Public Opinion Poll Results. In many countries there are consequently restrictions on the publication of opinion polls (Spangenberg 2003; Ferguson and de Clercy 2005; Donsbach and Hartung 2008), including regulations regarding what kind of methodological information the media should include every time they publish a poll.
In Sweden there are no government regulations (Petersson and Holmberg 1998; Petersson 2008), although the recommendations made by ESOMAR/WAPOR applies to the Swedish media as well as to the media in other countries. According to ESOMAR/WAPOR, when newspapers publish poll findings they should always be accompanied by a clear statement of (a) the name of the research organisation carrying out the survey, (b) the universe effectively represented, (c) the achieved sample size and its geographical coverage, (d) the dates of fieldwork, (e) the sample method used, (f) the method by which the information was collected, and (g) the relevant questions asked. The points (a) through (d) also applies to broadcast media.
Providing their users with this kind of information is one important means through which the media can provide people with the information they need to assess the value of published polls. What matters most from this perspective is thus not whether most news consumers actually are interested enough to use the methodological information to evaluate a poll, but whether the media give their consumers enough information to be able to do so. From this perspective, it is problematic that most studies have found that the media fail when it comes to providing the methodological information that ESOMAR/WAPOR recommends that the media publish (Welch 2002; Ferguson and de Clercy 2005;). As there are no reasons to expect the Swedish media to be better at providing this kind of information than the media in other countries, the fifth hypothesis is: When publishing polls during the election campaigns 1998-2006, the media most often failed to provide the methodological information recommended by organisations such as ESOMAR/WAPOR (H5).
A Matter of Focus
The promise of opinion polls is that they may "help bridge the gap between the people and those responsible for making decisions in their name" and that they - at least theoretically - can provide a "reliable measure of the pulse of democracy" (Gallup and Rae 1940, 14). By commissioning and covering opinion polls, the media can give voice to the people, and thus strengthen the role of the people in political communication and governing processes.
If polls have the potential to give voice to the people, this potential is nevertheless not realised automatically. It takes a well-designed poll to truly capture people's thoughts, opinions, or attitudes, with enough variables to allow at least somewhat sophisticated analyses (Bishop 2005). It also takes a journalistic interest in and focus on the people's response patterns, and a journalistic skill in interpreting people's responses and how they are correlated as well as how different response patterns can be explained (Meyer and Potter 2000). Even if these requirements were fulfilled, at the end of the day people have no say with respect to what the polls should focus on. Based on this notion, it has been argued that opinion polls largely serve the interests of political and media elites rather than the interests of the people (Bourdieu 1979). Opinion polls might furthermore produce artifacts as well as facts about public opinion, creating an illusion of public opinion where none exists (Bishop 2005). Thus, while the sheer publication of opinion polls might send a "continuous symbolic message that the opinions of 'everyone' matters" (Lavrakas and Traugott 2000, 10), it cannot be assumed that the media give voice to the people by commissioning and publishing opinion polls.
In the context of this paper, the main question then is whether the Swedish media during the election campaigns 1998-2006 mainly used opinion polls to give voice to the people - or rather as a means to serve the media's own need for compelling news narratives and for giving voice to the news journalists themselves. When reporting polls, did the media focus on the people whose opinions were polled y or on other actors in the political communication processes? More precisely, the main research question (Q1) is: When the media published opinion polls during the 19982006 campaigns, did they mainly focus on those polled and their opinions, or on aspects related to the media themselves or to political actors and their interrelationships and quest for public support?
Methodology and Data
To test the hypotheses above, this study utilises a quantitative content analysis of the two national public service television news shows Rapport and Aktuellt, the most important commercial television news show TV4 Nyheterna, the two major
national morning newspapers Dagens Nyheter and Svenska Dagbladet, and the two
national newsstand tabloids Aftonbladet and Expressen.
The time period for the study was the three weeks before the 1998, 2002 and 2006 Swedish national elections. The unit of analysis is news articles (newspapers) or thematic news stories (TV). For newspapers, the selection includes all articles published on the news pages in the main section, on pages with the vignette "politics" or its equivalent, or which in headlines, lead paragraphs or photos explicitly (^q referred to domestic political actors or institutions. Articles shorter than 10 lines are
not included. For television news, the selection includes all thematic news stories with the vignette "politics" or its equivalent, or which in words or images explicitly referred to domestic political actors or institutions. The full dataset consists of 1,022 news stories from the 1998 election, 1,154 news stories from the 2002 election, and 1,187 news stories from the 2006 election.
The coding sheet included a number of variables that will be used in this study. First, coders were instructed to code whether the news story presented results from one or several scientific opinion polls, that is, polls with a representative sample. Coders were instructed to code "yes" only if the presentation was rather complete. Brief references to one or several polls were not sufficient to be considered as presentations of polls. Second, if the news story presented results from at least one opinion poll, coders were instructed to code the main object of the poll. Four alternatives were available for coders: (1) vote intention or party preferences, (2) candidate or party leader preferences, (3) political issues, and (4) other. Third, if the news story presented results from at least one poll, coders were instructed to code whether the following methodological information was included in the presentation: (1) exact question wording, (2) number of respondents, (3) population, (4) margin of error or whether changes are outside of the margin of error, (5) method of collecting the survey responses, and (6) name of the organisation that sponsored the poll.
To capture the journalistic style of the news stories, coders were instructed to code whether the journalistic style was descriptive - told what happened in a rather straightforward style, or interpretive - analysed, evaluated, or explained a situation while also describing it. While all news stories contain both descriptive and interpretive elements, this variable was coded based upon the main story line or angle and whether the journalistic style largely was descriptive or interpretive.
The coding sheet also included several variables focusing on the framing of politics. Most importantly in this context, coders were instructed to code whether a (1) an issue metaframe or a (2) game metaframe was dominant in the news stories. Briefly, "game frame" refers to news stories that frame politics in terms of a game, personality contest, strategy, or personal relationships between political actors not related to issue positions. "Issue frame" includes stories that focus on issues and issue positions. News stories in which other frames were dominant will be treated as missing.
Unfortunately, for newspaper articles, coders coded headlines, lead paragraphs, main body and last paragraph separately, while for television news stories, they were coded in their entirety. To achieve comparability between newspaper articles and television news stories, in this study the codes for the main body of the newspaper articles will be used, disregarding the coding for the headlines and the lead and final paragraphs.
To answer RQ1, this study will take a closer look at the pattern of results revealed when testing the hypotheses above. In addition, a qualitative content analysis was performed, where all news stories with polls published in the largest morning newspaper Dagens Nyheter and the largest tabloid Aftonbladet during the 2006 campaign were carefully read several times to see in what ways - or if - these media focused on those polled or on other aspects related to the media themselves or to political actors an their interrelationships and quest for public support. Thus, RQ1 will be answered by combining the results from the quantitative content analysis with a qualitative content analysis.
Results
According to the first hypothesis, it can be expected that the number of news stories in which opinion polls were reported increased between the 1998, 2002 and 2006 election campaigns. As shown by table 1, this hypothesis receives some but not unequivocal support. Furthermore, the trend is not linear. The number of news stories with polls being reported shrank from 98 in 1998 to 88 in 2002, before increasing to 117 in 2006. There are also some differences between media types. Most polls were published in the newspapers, with television news being more restrictive. However, it should also be noted that the number of news stories with polls being reported went up in both public service television and commercial television news in 2006. In terms of percentages, the share of news stories with polls being reported has consistently been higher in commercial than in public service television news, while the differences between the tabloids and the morning newspapers have been minor. Overall, 9.5 percent of the news stories in 1998 reported results from opinion polls, while the corresponding shares in 2002 and 2006 were 7.6 and 9.9 percent respectively. In both raw numbers and share of news stories, 2002 thus appears to be an exception. Partly this can be explained by the fact that one of the tabloids, Expressen, in 2002 experimented with using real-time response measurements instead of polls to measure how the party leaders succeeded when being interviewed or debating on TV. If the publication of results from these experiments had been considered as polls reports, the number would have increased by seven.
Table 1: Frequency of Poll Reports in Swedish Election News, 1998-2006.
£ 3
a
0
e *
1 *
u 0 C
(ju
fN KD
	1998	2002	2006	N
Tabloids	50	28	38	1059
Morning press	36	48	55	1441
Public service TV	7	6	14	632
Commercial TV	5	6	10	231
Sum	98	88	117	3219
Whether it should be considered "beyond reason" (Patterson 2005) that about 10 percent of all election news stories typically include the presentation of opinion poll results is an open question. Any assessment of the media's use of opinion polls should go beyond the raw numbers, however, which leads to H2, predicting that most of the published polls during the 1998-2006 campaigns were about people's party or party leader preferences, i.e., that most polls were so called horse race polls.
The results show that most polls indeed were about people's party preferences or their perceptions of the party leaders and how they fared when being interviewed or facing off in debates on television, whereas polls on people's issue stands were less common - particularly in 1998. Thus, H2 receives support. In 2002 issue polls became more common however, and continued to be rather prominent in 2006. Partly this new emphasis on issue polls can be explained by an interest in public journalism (Strömbäck 2004), which preaches that the media should focus on the is-
sues people care most about rather than the issues the parties want to stress (Merritt 1998). Two other reasons for investigating and focusing on the issues people care CO most about might be: first, to legitimise the media's focus on some issues instead ^ of others, and second, to market the media as giving voice to the people.
Table 2: Contents of Opinion Polls Published in the Swedish News Media 19982006 (in percentages)
	1998	2002	2006
Party preferences	67.3	60.2	44.4
Candidate perceptions	20.4	15.9	21.4
Issues	7.1	19.3	24.8
Other	5.1	4.5	9.4
Sum	99.9	99.9	100
N	98	88	117
In 1998 only 7 percent of the published polls focused on issues, but in 2002 this share increased to 19 percent and in 2006 to 25 percent. At the same time the share of published polls focusing on people's party preferences or vote intention decreased from 67 percent in 1998 to 44 percent in 2006. The share of published polls focusing on people's preferences for the party leaders or their opinions on how these fared when being interviewed or facing off in televised debates decreased between 1998 and 2002, but rebounded in 2006. Hence, the increasing focus on people's issue preferences was mainly at the expense of polls on people's party preferences or vote intention. Aside from this it can be noted that the tabloids published almost all of the polls - 51 out of 59 - focusing on people's preferences for the party leaders or opinions on how they fared when being interviewed or debating on TV.
If both the polls focusing on people's party preferences and perceptions or evaluations of the party leaders are considered as horse race polls, then the share of horse race polls declined from about 88 percent in 1998 to 76 percent in 2002 and 66 percent in 2006. Thus, while horse race polls constitute a clear majority of all published polls, these results suggest that it would be misleading to assume that the media's focus on opinion polls automatically contribute the horse race reporting - at least in the Swedish case.
This leads to the third hypothesis (H3), stating that there will be a positive correlation between the publication of opinion polls and the framing of politics as a game. To test this hypothesis, separate analyses were run for newspapers and television news in each election cycle. News stories where an issue frame was dominant was coded as 1 and news stories where a game frame was dominant as 2, whereas 1=no opinion poll reported and 2=opinion poll reported. To predict the framing of politics from the publication of opinion polls, Somer's d was used.
The results show that there are consistent and positive correlations between whether the media published opinion polls and the framing of politics as a game rather than as issues (see Table 3). There is only one exception to this rule, related to television news during the 2002 election campaign. Out of the 12 poll reports in television news in 2002, five framed politics as issues whereas seven framed politics as a game.
Table 3: Correlation Between the News Media's Publication of Opinion Polls and the Framing of Politics as a Game (Somers' d).
	1998	2002	2006
Newspapers	.510***	.566***	.332***
TV	.538***	.257	.360**
** and *** indicate that the correlations are statistically significant at the .01 and .001 level respectively.
In general, the correlations were strongest in 1998 and, for newspapers, in 2002, and weakest although still significant in 2006. This is consistent with the results above, showing a decline in horse race polls and an increase in issue polls during the last two elections and particularly in 2006. Nevertheless, the results support H3 and those who claim that the media's focus on opinion polls contribute to the framing of politics as a game rather than as issues (Broh 1980; Patterson 2005; Rosenstiel 2005; Farnsworth and Lichter 2007).
The fourth hypothesis (H4) predicted that there would be a positive correlation between the publication of opinion polls and an interpretive - as opposed to a descriptive - journalistic style. This hypothesis was not supported, however, as only one significant correlation was found: In 2006, there was a significant and positive correlation (Somers' d = .237) between the publication of opinion polls and an interpretive journalistic style. In the other cases, neither positive nor negative significant correlations were present. The direction of the correlations was generally positive, but the correlations were not significant.
Turning to the quality of the media's reporting on opinion polls, H5 predicted that when publishing polls during the 1998-2006 election campaigns, the media most often failed to provide the methodological information recommended by organisations such as ESOMAR/WAPOR. To test this hypothesis, the presence of six types of methodological information was included, which broadly corresponds to the recommendations by ESOMAR and WAPOR and the type of methodological information that in some countries are regulated by government and in others through self-regulation (Welch 2002; Spangenberg 2003; Ferguson and de Clercy £	2005). As there are fundamental differences between newspapers and television
- most clearly seen in the lighter recommendations that ESOMAR/WAPOR apply to broadcast news compared to newspaper news - the presentation of the results for these two kinds of media are held separate. Table 4 shows the share of different types of methodological information that was present in the newspapers in each of the election campaigns, whereas table 5 shows the same information with respect to television news.
The results show that neither newspapers nor television news did a good job at jß providing the kind of methodological information that should be presented when a poll is published. On average, the newspapers published only about a third of the methodological information in 1998, and barely a majority in 2002 and 2006. Television news did an even worse job, on average publishing less than a quarter of all methodological information in all three elections. Although the recommendations directed at broadcast news typically is lighter than those directed at newspapers, it is noteworthy how seldom both newspapers and broadcast news (jq publish essential information such as the question wording, the population and
0
fi *
I
C >
Table 4: Percentage of Election News Stories in Newspapers and Television News Reporting Methodological Information in Poll Stories, 1998-2006.	lo
	1998		2002		2006	
	Newspapers	Television	Newspapers	Television	Newspapers	Television
Question wording	11.6	16.7	50.0	8.3	58.1	16.7
No of respondents	66.3	8.3	80.3	16.7	86.0	12.5
Population	12.8	16.7	47.4	16.7	45.2	45.8
Changes outside the margin of error	5.8	0	42.5	0	18.3	0
Interview method	12.8	0	25	16.7	38.7	0
Organisation that sponsored the poll	83.7	50.0	89.5	75.0	81.7	58.3
Mean	32.2	15.3	55.8	22.2	54.7	22.2
N	86	12	80	12	93	24
whether changes are within or outside of the margin of error. Although both the newspapers and the broadcast news, on the aggregate at least, have improved their reporting of methodological information across these three election campaigns, the overall picture is one of failures at providing people with the kind of information they need to evaluate a published poll and its quality. In this, the Swedish media appear to be neither significantly better nor worse than the media in other countries such as the U.S. (Welch 2002), Canada (Ferguson and de Clercy 2005) or Germany (Brettschneider 1997). All in all, although the newspapers in 2002 and 2006 published a bare majority of the methodological information, H5 is largely supported by the results.
While the results do not stand out in comparison with research on published methodological information in other countries, they are still noteworthy considering that "It is in the best interest of both newspapers and polling organisations to have newspapers [and broadcast news] disclose more information about polls than they are currently doing" (Welch 2002, 112). Although not all news consumers are interested in or have knowledge enough to use methodological information on polls, for media (and polling organisations) that need to be trusted sources of information, failing at providing easily accessible but essential information might hurt their credibility.
This might be particularly true if the media want their audiences to perceive them as using polls in order to give voice to and stand on the side of the people against those in power. In such a case, a prerequisite should arguably be that the media give the same people whose voices they purport to amplify the means to assess and evaluate the polls being reported on - particularly considering how easy it is to misuse polls (Bishop 2005) and that the media, in the Swedish case, are aware of the recommendations regarding what methodological information they should publish (Mellin 2002; Strömbäck 2004).
This brings us to the main research question (RQ1) in this study: When the media published opinion polls during the 1998, 2002 and 2006 Swedish election campaigns, did they mainly focus on those polled and their opinions, or on other aspects related to the media themselves or to political actors and their interrelationships and quest for political power?
To address this question, let us first consider the pattern of results so far. While the sheer number of published polls might suggest that the media have an interest in people's opinions, the results have shown that a majority of all polls focus on people's party preferences or evaluations of how the party leaders fared in televised interviews or debates. The most common question is whom people would vote for if the election was held today, in addition to questions on "who won" debates or how people would grade the party leaders' performances in major television interviews. Although the number of issue polls has increased, they still constitute a minority. Consequently, it is logical that the publication of opinion polls is consistently and positively correlated with the framing of politics as a game. In this, the critics of how the media use polls have it right; opinion polls spur horse race reporting. In this kind of reporting, almost all of the media's focus is on the candidates and what the polls might mean for their prospects on Election Day.
The results have also shown that while there is only one significant correlation between the publication of opinion polls and an interpretive journalistic style, this correlation was positive. Although the results suggest that it cannot be claimed that the publication of opinion polls predicts an interpretive journalistic style, it does not preclude that journalists often use opinion polls for analyses and interpretations; as noted by Lavrakas and Traugott (2000, 4), polls are attractive to journalists because these allow them "a quasi-objective, proactive role in the newsmaking process." In this context it is important to note that about three quarters of all poll reports in 2002 and 2006 (data not available for 1998) came out of polls commissioned by the media themselves. Finally, the results have also shown that the media fail when it comes to providing people with the kind of methodological information they need to evaluate the polls and their quality.
Taken together, these results suggest that the media mainly use polls to get access to unique and new information that can be used either to evaluate and analyse the horse race or - in the case of issue polls - to legitimise why the media focus on some issues rather than other. y	However, to get a thorough understanding of the media's use of opinion polls, a
closer reading is essential. Therefore, a qualitative content analysis of all poll reports in Aftonbladet and Dagens Nyheter in 2006 was performed. The main purpose of this analysis was to investigate to what extent and how the media, when publishing polls, focused on the people and their opinions, as opposed to the political actors or the media themselves.
The main result of this qualitative analysis is that the people is strangely absent
, , , , _
from almost all poll results in the investigated media. Most of the time the media only report the raw numbers or tendencies that can be found in the polls, and then either settle with this - in some, but rather few, cases also looking at differences
this information to analyse the political game or to pose questions to political candidates. In fact, the only time one of these media - Aftonbladet - explicitly invoked "the people" when reporting a poll was after then - Prime Minister Göran Persson was interviewed on TV. The poll showed that Persson only received 3.4 on a scale from 1-5. When Persson dismissed this result, Aftonbladet headlined the article: "Perssons insult to the viewers." The viewers were here used as representatives for (^q the people, and when Persson dismissed the poll result, he scoffed at the people.
Interesting to note in this context is that the poll in question was based on only 370 respondents who had watched the whole or parts of the interview; how many had seen the whole interview was not reported. This N is indeed very low for inferences about what the people or the viewers thought in this matter.
This example also shows how the media use poll results, when they do not just report the results without analysing or using them further: they use the polls to evaluate and pose more or less critical questions to the parties or their representatives. There is hardly anything in the coverage of these poll results that suggest a genuine interest in what the people think and why they think as they do.
The same is more or less true with respect to the media's publications of issue polls. Although the media seldom use issue polls to evaluate the parties - if an issue is not included in a referendum - they use them instead to legitimise why they focus on some issues instead of others. In both cases, the media use the polls as a tool for serving their own purposes, rather than to give voice to the people. In most cases, the media furthermore report the results without making it clear whether changes are significant or not and what questions and response alternatives were used. Even when reporting issue polls, the media seldom focus on the people and their thoughts or opinions. At best, the media present how the results differs between demographic groups, but beyond that, the people are strangely absent considering that their opinions are supposed to be the foundation of the polls being reported.
Hence, the results from both the quantitative and the qualitative content analysis suggest that the media do not use polls mainly to give voice to the people, but rather to serve the media's own needs of unique news stories and of information that can be used to inform their horse race coverage, the framing of politics as a game, and as a help when deciding on and legitimising their issue coverage. To conclude: If polls indeed are to be perceived as the pulse of democracy, the power lies with those who measure the pulse, not with those whose pulse is being measured.
Discussion and Conclusions
The main purpose of this study was to investigate how the Swedish media during the last three election campaigns used opinion polls and whether or not the media ultimately used opinion polls in order to give voice to the people - or to the media and the journalists themselves. As hypothesised, the results suggest an increase in the number of news stories in which opinion polls were reported. Most of the polls were about either people's party preferences or their evaluations of the party leaders, i.e., how they fared when being interviewed or when they debated on television. Issue polls have become more common, but horse race polls dominate. As hypothesised, the results also show consistent support for the notion that the publication of opinion polls spurs the framing of politics as a game. However, the hypothesis that there is a correlation between the publication of opinion polls and an interpretive journalistic style was rejected. With respect to the quality of opinion poll reporting, the results show that the media largely failed in providing people with the kind of methodological information needed to assess the poll results and the quality of the polls. The only good news in this respect is that the media have improved somewhat since 1998.
Normatively, the most important question is however related to whether or not the media mainly used opinion polls to give voice to the people. Here the results
from both the quantitative and the qualitative content analysis suggests that the media seldom use opinion polls in ways that suggest a genuine interest in the people and their thoughts and opinions. Rather, the main reasons for the media's interest in opinion polls appear to be: First, to get exclusive news and unique information; second, to have information that can be used when analysing the political horse race; third, to have information that can be used when posing more or less critical questions to political actors; fourth, to have information that allow journalists a quasi-objective, proactive role in political communication processes; fifth, to send a symbolic message that the media care about ordinary people and their opinions, and sixth, to legitimise why the media focus on some rather than other issues.
Taken together, the results of this study thus show that if the media indeed use opinion polls to give voice to the people, they fail. As used by the media, opinion polls very seldom serve as vox populi. Rather, opinion polls serve as vox media. Even if we assume that opinion polls do a good job at measuring public opinion, which often is highly questionable, the media do a poor job at using opinion polls to give voice to the people.
As least this holds true with respect to the Swedish media's use of opinion polls in the last three election campaigns. Whether it also holds true in other countries and in the media's use of opinion polls between election campaigns remains to be investigated. Although most research suggests that the media in Sweden neither do a better nor a worse job than the media in other countries (for an overview, see Brettschneider 2008), there is a noticeable lack of comparative research on the media's use of opinion polls. This calls both for caution when drawing conclusions and for more comparative research in this area. Only by more comparative research, both across countries and across time, will it be possible to find generally valid knowledge regarding whether opinion polls, as covered by the media, mainly serve as vox populi or vox media.
References:
ÜAtkin, Charles K. and James Gaudino. 1984. The Impact of Polling on the Mass Media. Annals of the
American Academy of Political and Social Science 472, 1, 119-128. Bauman, Sandra L. and Paul J. Lavrakas. 2000. Reporters' Use of Causal Explanation in Interpreting Election Polls. In P. J. Lavrakas and M. W. Traugott (eds.), Polls, the News Media, and Democracy, 162-181. New York: Chatham House.
a	Bishop, George F. 2005. The Illusion of Public Opinion. Fact and Artifact in American Public Opinion Polls.
Lanham: Rowman & Littlefield. Bogart, Leo. 2000. Politics, Polls, and Poltergeists: A Critical View of the 1996 Election. In P. J. Lavrakas & M. W. Traugott (eds.), Polls, the News Media, and Democracy, 301-318. New York: Chatham House.
Bourdieu, Pierre. 1979. Public Opinion Does Not Exist. In A. Mattelart and S. Siegelaub (eds.),
Communication and Class Struggle, 124-130. New York: International General. Brettschneider, Frank. 1997. The Press and the Polls in Germany, 1980-1994. Poll Coverage as an Essential Part of Election Campaign Reporting. International Journal of Public Opinion Research 9, 3, 248-265.
S
Traugott (eds.), The Sage Handbook of Public Opinion Research, 479-486. London: Sage.
" ' " ' ~ " ~ ' -........ " '
Brettschneider, Frank. 2008. The News Media's Use of Opinion Polls. In W. Donsbach and M. W.
00
Broh, C. Anthony. 1980. Horse-Race Journalism: Reporting the Polls in the 1976 Presidential Election.
Public Opinion Quarterly 44, 4, 514-529. Cappella, Joseph N. and Kathleen Hall Jamieson. 1997. Spiral of Cynicism. The Press and the Public
(^q	Good. Chicago: University of Chicago Press.
De Vreese, Claes H. and Matthijs Elenbaas. 2008. Media in the Game of Politics: Effects of Strategic
Metacoverage on Political Cynicism. The International Journal of Press/Politics 13, 3, 285-209.	^
Donsbach, Wolfgang and Uwe Hartung. 2008. The Legal Status of Public Opinion Research in the	*>0
World. In W. Donsbach and M. W. Traugott (eds.), The Sage Handbook of Public Opinion Research, 431-440. London: Sage.
Donsbach, Wolfgang and Michael W. Traugott. 2008. Introduction. In W. Donsbach and M. W. Traugott (eds.), The Sage Handbook of Public Opinion Research, 1-5. London: Sage.
ESOMAR/WAPOR. 2008. Guide to Opinion Polls Including the ESOMAR International Code of Practice for the Publication of Public Opinion Poll Results.
Farnsworth, Stephen J. and S. Robert Lichter. 2007. The Nightly News Nightmare. Television's Coverage of U.S. Presidential Elections, 1988-2004 (2nd Edition). Lanham: Rowman & Littlefield.
Ferguson, Peter A. and Cristine de Clercy. 2005. Regulatory Compliance in Opinion Poll Reporting During the 2004 Canadian Election. Canadian Public Policy 31, 3, 243-257.
Gallup, George and Saul F. Rae. 1940. The Pulse of Democracy. The Public Opinion Poll and How It Works. New York: Simon & Schuster.
Glasser, Theodore L. and Charles T. Salmon. 1995. Public Opinion and the Communication of Consent. New York: The Guilford Press.
Gollin, Albert E. 1980. Exploring the Liason Between Polling and the Press. Public Opinion Quarterly 44, 4, 445-461.
Herbst, Susan. 1995. On the Disappearance of Groups: 19th and Early 20th Century Conceptions of Public Opinion. In T. L. Glasser & C. T. Salmon (eds.), Public Opinion and the Communication of Consent, 89-104. New York: The Guilford Press.
Herbst, Susan. 1998. Reading Public Opinion. How Political Actors View the Democratic Process. Chicago: University of Chicago Press.
Hoffman, Nicholas von. 1980. Public Opinion Polls: Newspapers Making Their Own News? Public Opinion Quarterly 44, 4, 572-573.
Holmberg, Sören. 2008. Opinionsmätningar tar plats. In O. Petersson (ed.), Medierna: folkets röst?, 137-167. Stockholm: SNS Förlag.
Holmberg, Sören and Olof Petersson. 1980. Inom felmarginalen. En bok om politiska opinionsundersökningar. Stockholm: Publica.
Ismach, Arnold H. 1984. Polling as a News-Gathering Tool. Annals of the American Academy of Political and Social Science 472, 1, 106-118.
Iyengar, Shanto, Helmuth Norpoth and Hahn S. Kyu. 2004. Consumer Demand for Election News: The Horserace Sells. Journal of Politics 66, 1, 157-175.
Kovach, Bill and Tom Rosenstiel. 1999. Warp Speed. America in the Age of Mixed Media. New York: The Century Foundation Press.
Ladd, Everett Carl. 1980. Polling and the Press: The Clash of Institutional Imperatives. Public Opinion Quarterly 44, 4, 574-584.
Lang, Gladys Engel and Kurt Lang. 1980. Polling on Watergate: The Battle for Public Opinion. Public Opinion Quarterly 44, 4, 530-547.
Lavrakas, Paul J. and Michael W. Traugott. 2000. Why Election Polls Are Important to a Democracy: An American Perspective. In P. J. Lavrakas and M. W. Traugott (eds.), Polls, the News Media, and Democracy, 3-19. New York: Chatham House.
Lewis, Justin. 1999. The Opinion Poll as a Cultural Form. International Journal of Cultural Studies 2, 2, 1999-221.
Mellin, Lena. 2002. Aftonbladet och opinionsmätningarna. In J. Strömbäck (ed.), Medierna och opinionsmätningarna, 18-23. Sundsvall: Demokratiinstitutet.
Merritt, Davis. 1998. Public Journalism and Public Life. Why Telling the News Is Not Enough. 2nd Edition. Mahwah: Lawrence Erlbaum Associates.
Meyer, Philip and Deborah Potter. 2000. Hidden Value: Polls and Public Journalism. In P. J. Lavrakas and M. W. Traugott (eds.), Polls, the News Media, and Democracy, 113-141. New York: Chatham House.
Patterson, Thomas E. 1993. Out of Order. New York: Vintage.
3 ft
O
(I)
o
C
(B
o
Patterson, Thomas E. 2005. Of Polls, Mountains. U.S. Journalists and Their Use of Election Surveys.
Public Opinion Quarterly 69, 5, 716-724. Petersson, Olof (ed). 2008. Medierna: folkets röst? Stockholm: SNS Förlag.
Petersson, Olof and Sören Holmberg. 1998. Opinionsmätningarna och demokratin. Stockholm: SNS Förlag.
Petersson, Olof, Monika Djerf-Pierre, Sören Holmberg, Jesper Strömbäck and Lennart Weibull. 2006.
Media and Elections in Sweden. Stockholm: SNS Förlag. Rosenstiel, Tom. 2005. Political Polling and the New Media Culture: A Case of More Being Less. Public
Opinion Quarterly 69, 5, 698-715. Salmon, Charles T. and Theodore L. Glasser. 1995. The Politics of Polling and the Limits of Consent. In T. L. Glasser and C. T. Salmon (eds.), Public Opinion and the Communication of Consent, 437-458. New York: The Guilford Press. Spangenberg, Frits. 2003. The Freedom to Public Opinion Poll Results. Report on a Worldwide Update.
Amsterdam: Foundation for Information. Splichal, Slavko. 2008. Fran opinionsstyrd demokrati till globala styrelseformer utan opinion In O.
Petersson (ed.), Medierna: folkets röst?, 29-50. Stockholm: SNS Förlag. Splichal, Slavko. 1997. Political Institutionalisation of Public Opinion Through Polling. Javnost - The Public 4, 2, 17-38.
Strömbäck, Jesper. 2004. Den medialiserade demokratin. Om journalistikens ideal, verklighetoch makt. Stockholm: SNS Förlag.
Strömbäck, Jesper and Lynda Lee Kaid. 2008. The Handbook of Election News Coverage Around the
World. New York: Routledge. Traugott, Michael W. 2004. Do Polls Give the Public a Voice in a Democracy? In M. A. Genovese and M. J. Streb (eds.), Polls and Politics. The Dilemmas of Democracy, 77-94. New York: State University of New York Press.
Traugott, Michael W. 2005. The Accuracy of the National Preelection Polls in the 2004 Presidential
Election. Public Opinion Quarterly 69, 5, 642-654. Weimann, Gabriel. 1990. The Obsession to Forecast: Pre-Election Polls in the Israeli Press. Public
Opinion Quarterly 54, 3, 396-408. Welch, Reed L. 2002. Polls, Polls, and More Polls: An Evaluation of How Public Opinion Polls Are
Reported in Newspapers. Harvard International Journal of Press/Politics 7, 1, 102-114. Wichmann, Wolfgang. 2008. Quality of Newspaper Reporting on Political Polls: A New Approach. Paper presented to the International Communication Association, Montreal, Canada,(May 2008.