The news media are generally assumed to be an important factor in election campaigns. Journalists, academics, and politicians routinely cite coverage as influential in the rise and fall of leaders, the polling performance of parties, the emergence of par- ticular issues during campaigns, even the outcomes on voting day. But media influence is more often presumed than demonstrated, and there is a relatively limited body of empirical work on the role of media in Canadian elec- tion campaigns.

To help fill this gap, McGill University’s Observatory on Media and Public Policy analyzed the contents of seven major Canadian dailies for the duration of the 2004 federal election campaign. Our goal was to track media cover- age in real time and to examine more general trends in the triangular relationship between media, the public, and politics. This project gave us the opportunity to investigate one of the critical questions in electoral studies: who is leading the campaign " the public, parties and politicians, or the media?

The answer to this question can change on a daily basis. One day, party press releases will dominate the head- lines; the next day, journalists may uncover a politician’s questionable past. So an election inevitably involves ongoing, multi-directional interac- tions between media, public and politicians " not unlike non-election periods, though usually at a much more heightened pace.

Our study focused on the role of newspapers during the election cam- paign, and attempted to shed light on whether this medium more often reflected or affected the campaign itself. Our preliminary conclusions suggest that, in the 2004 federal elec- tion campaign at least, Canadian media played the role of follower rather than leader. To the extent that this trend is generalizable, this has implications for our overall view of the role of the Canadian media in the political process.

The recognition that the press was primarily a follower also has important implications for what we identify as the failures or lacunae of the 2004 campaign. Foremost among these failures, we suggest, are (1) the predominance of horserace coverage during that campaign, and (2) the relative lack of issue-oriented stories in the 2004 campaign. The two are connected and both are clearly evi- denced in our data on media coverage of the election.

Starting a week before the official election campaign kick-off, the Observatory’s team of coders began to scrutinize the main news sections of seven major Canadian dailies: The Globe and Mail, National Post, Toronto Star, Calgary Herald, Vancouver Sun, La Presse and Le Devoir. Each of the seven coders was responsible for a different newspaper each week so that we could test for any coder effects or bias. Coders noted all articles relating to federal politics, including reportage, analysis, opinion, and editorials. This content was coded for mentions of issues, parties, and leaders, as well as positive, negative or neutral tone. The precedence of these mentions and their prominence within the paper was also noted, along with other factors. Coding began on Monday, May 17, and continued until the day of the election, June 28. In total, 2,948 arti- cles were analyzed " 1,839 news sto- ries and 1,089 editorial and opinion pieces. (Detailed data are available at

This method of coding indicated which parties or leaders received posi- tive or negative treatment, and which issues dominated the campaign, for instance. It also showed which parties were associated (positively or nega- tively) with leading issues, and sug- gested which parties were driving coverage of the campaign or respond- ing to the initiatives of their oppo- nents. The coding also indicated how treatment of parties, leaders and issues varied across newspapers, and across regions in Canada.

All of the newspaper articles were coded as either ”œissue” or ”œhorser- ace/campaign” focused, and the results are presented by week in figure 1. The volume of horserace cover- age is striking: after a brief period of policy discussion in week two of the cam- paign, horserace coverage increased gradually to over 60 percent of all election articles. The space left for substantive policy discussion was, accordingly, relatively limited.

Figure 2 nonetheless illustrates those issues that were discussed most in the campaign. Government accountability, including the sponsorship scandal, was the clear winner in the early days of campaign. Leadership and accountability were obviously important concerns to vot- ers, but we noted at the time the need for a discussion of actual policies. Accountability is important, but is not exactly a policy mandate.

We accordingly noted the emerg- ing salience of health care and taxes in the second week of the campaign. The trend began with Liberal leader Paul Martin’s introduction of the Liberal party’s health care package and continued with all of the parties’ positions on taxes. The result was what appeared at the time to be a classic and important competition between taxes and spending. It was a debate similar to and perhaps fore- shadowed by the recent provincial election in Ontario in the fall of 2003.

Health care and taxes played a more limited role for the rest of the campaign, however. By week three, press coverage was dominated by Liberal criticism of Conservative social policies. Same sex marriage and abortion policy dominated headlines during this week; the fol- lowing week saw the re-emergence of accountability. So while weeks one and two focused primarily on the parties’ policy proposals, subse- quent weeks shifted to reflect the tone of partisan attacks " the Liberals discussing (potential) Conservative social policies, and the Conservatives (with help from other parties) focusing on recent Liberal accountability scandals.

This is apparent if we look at the relative salience of issues in isolation. Note that figure 1 shows declining issue coverage in the latter half of the campaign, so that the issues tracked in figure 2 actually make up a declining share of total election coverage. Overall, then, the issues gap in the campaign was the product of the replacement of party proposals with party attacks, and the more general decline in both forms of issue cover- age in the final weeks of the cam- paign. Policy issues did play a role in the campaign, but that role was limit- ed to start with, and increasingly lim- ited over time.

The Observatory’s coding grid included the first three political par- ties and the first three party leaders’ names mentioned in each article. In addition, each party and name was coded for tone " positive, negative or neutral. The decision rule was that unless the mention was obviously and intentionally positive or nega- tive, a mention was neutral. For instance, reporting a Harper speech in which the Conservative leader object- ed to or attacked something about Paul Martin was considered neutral " just reporting the news. Reporting that speech and using it to further discuss Martin’s failings was consid- ered negative, however.

The overall result was that men- tions in news stories were pre- dominantly neutral, and mentions in editorial and opinion pieces were mainly negative or positive. To the extent that there was bias in news cov- erage, it remained relatively slight. Not unsurprisingly, opinions were more frequently expressed in the editorial pages.

Figure 3 shows results for ”œnet coverage” (percent positive men- tions minus percent negative men- tions) across newspapers. The general trend is evident " in every paper, the proportion of positive or negative coverage was much greater in opinion pieces than in news sto- ries. There are significant differences across newspapers, however " and in several cases the bias evident in opinion pieces seemed to seep into news coverage. This was noted in the coding of the Calgary Herald, and to some extent of the Toronto Star. The National Post published a num- ber of pro-Conservative editorials, but news coverage appeared relative- ly neutral. However, it is not insignificant that of the seven dailies tracked by the Observatory, the Post is the only newspaper that printed more opinion pieces about the election (212) than news stories (151). This ratio of news to opinion was the inverse of what we noted in the remaining six newspapers in our sample. Each of the other papers ran substantially more news articles than opinion pieces. Given the dominance of opinion in its pages, the contents of the Post were conse- quently relatively pro-Conservative. 

The issue of media bias is seldom hotter than during an election campaign, although the data we collected speak to this issue only in part. The trends identified here reflect the volume of negative or positive coverage of parties during the campaign " whether they reflect bias is another matter, however. While the word ”œbias” has negative connotations, partisan differences across newspapers can be seen to represent a healthy diversity of opinion among media outlets. (Though note that this diversity argument is dependent on there being a plurality of newspapers within each regional market.) In any case, we make no claims of bias here. We do note, however, that differences in the tone of news coverage do exist across news- papers, and " where trends in tone are evident " these trends do tend to match the partisanship reflected in editorial and opinion pieces.

More important still is the evo- lution in the tone of coverage over the course of the campaign. Figures 4 and 5 show ”œnet coverage” meas- ures for each of the major parties and leaders throughout the cam- paign. Most notable in figure 5 are the punctuations in the series around the time of the leaders’ debates. Following the French lan- guage debate on June 14, Gilles Duceppe’s ”œnet coverage” (percent positive mentions minus percent negative mentions) increased from 0 to +15. The Bloc Québécois leader was clearly the winner, at least where tone of news coverage is the measure. The other beneficiary of the French-language debate was Liberal leader Paul Martin, whose net tone moved up almost as much, albeit starting at a much lower point (-12 to 0). Conservative leader Stephen Harper’s coverage was essentially unchanged while NDP leader Jack Layton’s net tone slid from -8 to -10. (Note that, for the sake of visual clarity, figures 4 and 5 show five-day averages rather than daily figures; these punctuations are consequently smoothed somewhat.)

These trends continued in the press coverage of the June 15 English- language debate, enhanced by an increase in horserace coverage from 50 percent to 60 percent of all election articles during the same week. Duceppe’s coverage remained positive; Harper’s coverage was generally neu- tral; Layton’s coverage was slightly more negative; and Martin’s net tone continued to rise. In fact, the Wednesday following the debates was the first time since the campaign began that Martin’s coverage was posi- tive rather than negative. As far as media were concerned, the Liberal leader was in ascendancy while Harper and Layton were falling.

This changed dramatically the next day. Coverage of leaders on Thursday, June 17 returned almost to the way it was before the debates. In particular, this meant a sharp drop in the tone of Martin’s mentions while his adversaries improved substantially.

Why did this happen? We suggest that the Ipsos poll, released 24 hours after the English-language debate, was critical. The poll showed that voters did not share media assessments of leader performance " voters were rel- atively less impressed with Martin, and more approving of Harper. Journalists appear to have corrected their initial impressions in light of public opinion.

Another opinion-driven shift in media content is apparent in the series tracking the order of party and leader mentions in articles. In the first week of the campaign, 61 percent of these articles mentioned Paul Martin first. In the second and third weeks, the proportion of Martin’s first mentions in election articles dropped to 44 and then 32 percent. Over the same period, Harper’s portion of first mentions moved from 19 percent in week one, to 37 percent in week two, to 49 per- cent in week three.

The trend reflects a change in the way in which the election was being framed. For the first week, the cam- paign was essentially about Paul Martin and the Liberal Party. Would Martin be able to deal with the spon- sorship scandal? What were his major policies? How were other parties react- ing to the Martin campaign? These were the questions that dominated coverage " coverage of a campaign that was essentially about Martin win- ning or losing.

That changed with the release of the Ipsos poll at the beginning of the second week, a poll suggesting that the Conservatives were making con- siderable gains in Ontario. Media dis- cussion of minority governments consequently increased, as did cover- age of Harper. The focus of campaign coverage widened somewhat, prima- rily in reaction to public opinion polls.

Trends in story first mentions reversed in the second half of the 2004 campaign. Martin’s share of first mentions increased by 12 per- cent from week three to week four and a further 4 percent going into week five. Harper’s first mentions tapered off by about the same amounts. These final two weeks were challenging for Harper and the Conservatives. The Conservatives appeared to stall in the polls, and the party made several missteps. First, the release of strongly worded state- ments linking Martin with lax child pornography laws caused a stir, and second, the Conservatives’ controver- sial plans for Air Canada became public. Harper’s decline in first mentions and tone of coverage (see figure 5) reflected these mis-steps.

The picture that emerges from coverage of leaders, parties an issues is one in which media were following much more than leading. Most of the dynamics in figures 1 through 6 can be explained by a combination of party-driven campaign events and shifts in public opinion that in each case preceded shifts in media cover- age.

Trends in issue coverage can be explained in large part by party press conferences and releases. The promi- nence of health care in week three fol- lowed the Liberal Party’s health policy statements the preceding Friday. The rise in coverage on taxes similarly fol- lowed the Conservative Party’s policy statements on the same day. Social housing rose in prominence due to an angry attack by the NDP’s Jack Layton. And the salience of government accountability was directly linked to Conservative, NDP and BQ press releases in the first part, and the latter half, of the campaign. Indeed, there were almost no trends in issue cover- age that were not the product of party communications.

Coverage of leaders and parties was not very different. We have already outlined the way in which coverage appears to have adjusted to opinion following the debate " again, a story of following rather than lead- ing. And this was the only really serous shift in tone during the cam- paign. Where first mentions are con- cerned, the shift from Martin- to Harper-focused coverage was largely the consequence of public opinion polls. (Note that these trends suggest that the importance of polls may lie in their effect on the nature of campaign coverage, rather than their direct effect on voters.)

There are, then, relatively few shifts in coverage over the campaign that were not relatively clearly pre- cipitated by campaign events or public opinion. The predominant role of media in the 2004 campaign, accordingly, seems to have been reflecting rather than leading the campaign. We do not suggest here that this is either good
or bad. We do suggest that it has important implications for why there was a lack of substantive issue content in the 2004 campaign.

In short, and contrary to what many observers and pundits have charged, the relative absence of policy issues was not primarily the fault of the media. If the campaign lacked in substance, it appears here as though the parties should take a good share of the responsibility. After a first week of substantive issues, attention turned away from the policies that par- ties stood for, and toward the issues that made other parties look bad. There was little policy discussion late in campaign. These trends were the consequence of party strategies.

Newspapers do have to write about an election, however, and the saving grace in 2004 was the tight race between the Liberals and Conservatives. This was the first election since 1988 for which the outcome was uncertain, and one of the few that could reasonably fuel speculation about the intricacies of a minority government. The horser- ace, it turned out, was one of the very few elements of this campaign that was interesting. It is no wonder that the media seized on this. And it is no coincidence that the declining trend in horserace coverage was reversed as Conservative fortunes improved and as party issue state- ments declined. 

Admittedly, conclusions about responsibility for the relative lack of issue coverage must also consider the parties’ expectations of media going into the campaign. Who is winning and who is losing; which leaders deal ”œknockout punches” in debates; and who can be expected to win the big prize on election night are all straight- forward, entertaining, and easy to package in three hundred words or less. Party strategists likely adjust com- munications accordingly " expecta- tions about media performance may therefore affect campaigns. These data cannot reveal how parties adjusted strategy to appeal to media. The relative culpa- bility of parties and media for the 2004 campaign is consequently perhaps not easy to attribute.

Our data are clear where prominence, tone and temporal precedence are concerned, however. The 2004 campaign was dominated by horserace over issue coverage, and even issue coverage focused increasingly on partisan attacks over substantive policy pro- posals. Negative coverage dominated positive coverage, particularly for Martin and the Liberal party. And when shifts in coverage occurred, these appear to have been driven by parties and public opinion rather than proactive journalists. This is important for future discussions about the role of media in election communication, the possible func- tions of a more proactive ”œpublic journalism,” and the behaviour of parties in Canadian elections.