BACK TO THE FUTURE: PRESS COVERAGE OF THE 2008 CANADIAN ELECTION CAMPAIGN STRIKES BOTH FAMILIAR AND UNFAMILIAR NOTES

Page created by Allen Reed
 
CONTINUE READING
BACK TO THE FUTURE:
PRESS COVERAGE OF
THE 2008 CANADIAN ELECTION
CAMPAIGN STRIKES BOTH
FAMILIAR AND UNFAMILIAR
NOTES
Blake Andrew, Lori Young and Stuart Soroka

The Media Observatory at the McGill Institute for the Study of Canada produced
another round of election news analysis for the 2008 federal election. Drawing on this
body of data and using a fully automated content analytic procedure, this article assesses
election news articles for prominence and tone of party and leader coverage, as well as
policy content. The main conclusions are three-fold. First, this campaign was increasingly
driven by the global and the Canadian economy. The environment, poised to control
the agenda, was overshadowed by the financial crisis in the United States. Second, the
Liberal Party — and particularly their new leader Stéphane Dion — was much less
prominent in this campaign than in previous elections. Third, while the Conservative
Party clearly received more coverage, it did not succeed in getting better coverage.

L’Observatoire des médias de l’Institut d’études canadiennes de McGill a produit
pour les élections fédérales de 2008 une autre analyse de la couverture de la
campagne. Sur la base des données recueillies, et en utilisant maintenant une
procédure automatisée d’analyse de contenu, les auteurs évaluent le ton ainsi que le
contenu politique des articles et la place consacrée à chaque parti et aux chefs. Ils
en tirent trois conclusions. Premièrement, la campagne a été de plus en plus centrée
sur l’économie mondiale et canadienne : l’environnement, dont on croyait au début
qu’il s’imposerait comme thème central, a finalement été éclipsé par la crise
financière aux États-Unis. Deuxièmement, la présence du Parti libéral, et notamment
de son nouveau chef Stéphane Dion, a été beaucoup plus discrète que dans les
campagnes précédentes. Troisièmement, le Parti conservateur a fait l’objet d’une
couverture plus importante sans qu’elle soit nécessairement plus favorable.

I
     t has been 20 years since economic issues were at the            main parties or leaders quite seemed prepared for. But per-
     forefront of a Canadian federal election campaign. In            haps the definitive contrast between 1988 and 2008 is that
     1988, the debate on free trade with the United States            the media spotlight on the Liberal Party (and particularly its
accounted for most of the policy coverage reported by                 leader) had faded considerably.
media. Since then federal elections have been mainly driv-                 As we shall see, though, the tone of this campaign was also
en by conversations about government accountability,                  markedly different. In media coverage, Stephen Harper and the
national unity and the enduring frame of a strategic “horse           Conservatives may have prevailed in this campaign partly as a
race” between parties and leaders. One of the key differences         function of receiving significantly more press coverage than
between 1988 and 2008 is that none of the main party lead-            the other parties and party leaders. But their success media-wise
ers in this campaign was prepared for the almost complete             seems to have been limited: coverage of the Conservatives was
domination of economic issues. Indeed, it became a cam-               increasingly negative relative to that of their opponents.
paign about the economy to an extent that none of the                 Negative coverage for the Conservatives was apparently not so

                                                                                                                     POLICY OPTIONS       79
                                                                                                                          NOVEMBER 2008
Blake Andrew, Lori Young and Stuart Soroka

     bad, though; and the sheer volume of        picture of Canadian newspapers’ cov-        framed, and point to which party or
     Conservative coverage overwhelmed the       erage of the 2008 election.                 leader is effectively driving coverage.
     other parties in a way that certainly has        The study includes every story
     not been evident for any party in the
     past two elections. One challenge ahead
     for all parties and leaders may be how to
                                                 from the main news sections of each of
                                                 the seven dailies that mentions the fed-
                                                 eral election or any of the parties or
                                                                                             T  he automated content analysis
                                                                                                thus tracks the volume and
                                                                                          prominence of coverage allocated to
     respond to and prepare for this peculiar-   party leaders. In total, just over 4,050 the main parties and leaders, as well as
     ly Conservative- and Harper-focused         articles were analyzed including news    trends in issue salience, efficiently and
     media environment.                          stories as well as editorial and opinion systematically. It also includes one
          The 2008 federal campaign offi-        pieces. The content analysis captures    final set of codes for the tone of cover-
     cially kicked off on the morning of         party, leader and issue prominence, as   age of parties and leaders. The auto-
                                                                                                       mated method we use to
     The sheer volume of Conservative coverage overwhelmed the analyze tone is well estab-
     other parties in a way that certainly has not been evident for                                    lished in social science liter-
                                                                                                       ature. Tone of coverage is
     any party in the past two elections. One challenge ahead for
                                                                                                       determined by the relative
     all parties and leaders may be how to respond to and prepare frequency of a pre-defined
     for this peculiarly Conservative- and Harper-focused media                                        list of sentiment-bearing
     environment.                                                                                      words in a selection of text.
                                                                                                       Our tone dictionary is
     September 7. Our analysis of media       well as the tone of coverage about par-     derived from several of the most well-
     content begins several days earlier, on  ties and leaders. For each newspaper, we    established and comprehensive lexical
     September 3, during the period when      use a computer-based procedure that         resources available for sentiment
     election speculation in Ottawa began     relies largely on simple text-based         analysis. The dictionary is composed
     in earnest. The study, then, spans the   searches.                                   of an expansive list of words (includ-
     entire campaign — for 42 days in total        Topic coding reflects the issue        ing inflected word forms and common
     — concluding on October 14, election     salience or policy focus of a cam-          idioms), all annotated for positive or
     day. This is the third instalment of our paign. In previous studies, human           negative valence. Sentiment classifica-
     federal election content analysis,       coders manually recorded the first          tion is much more challenging than
     extending our previous efforts in 2004   three topics mentioned in each arti-        identifying topic and actor mentions,
     and 2006. While analysis of past cam-    cle. Here, the automated procedure          of course, since tone tends to be more
     paigns was performed manually by         calculates topic codes using a diction-     dependent on the relation between
     human coders actively reading the        ary of keywords for each major policy       words. To determine the tone about
     newspapers, the 2008 analysis relies     domain. If an article mentions at least     particular actors, then, we apply sim-
     entirely on a new automated content      three keywords for a given policy           ple lexical rules to account for the
     analytic procedure.                      topic, we assume the article has some       proximity of keywords in the tone dic-
          The automated method relies on      content dealing with that topic. Based      tionary to actor references. This
     dictionaries of categorical keywords,    on this procedure, then, our election       requires some discussion, of course, a
     which are currently available only in    issue analysis reflects the percentage      brief version of which we include here.
     English. As a consequence, our 2008      of articles that discuss, at least in pass- (More detailed methodological infor-
     study covers only English-language       ing, a given topic.                         mation is available at http://media-
     press. The sample includes the same           The study also tracks the total        observatory.mcgill.ca.)
     English-language newspapers as in        number of mentions of all major par-             The automation procedure for tone
     2006 and 2004, The Globe and Mail,       ties and leaders, as well as the place-     begins by identifying all sentences in
     National Post, Toronto Star, Vancouver   ment of each mention. To identify the       which any of the major parties or lead-
     Sun and Calgary Herald. To that group,   first party or leader mentioned in an       ers is mentioned and then counting the
     we have added the Montreal Gazette       article, we automatically count charac-     number of positive and negative words
     and the Ottawa Citizen, which replace    ters from the beginning of the article,     in those sentences. To generate the
     Le Devoir and La Presse in terms of vol- directly replicating the manual coding      measure of “net tone,” by article, we
     ume though clearly not in terms of       of first mentions in our previous stud-     subtract the number of negative words
     content. There are some gaps in          ies. First mentions provide valuable        from the number of positive words,
     regional coverage, then, and we now      information about the relative salience     taken as a proportion of all words in
     lack French-language Quebec media        of parties and party leaders during the     sentences mentioning a given party or
     entirely. Our results should thus be     campaign; they are indicative of the        party leader. The result is a measure of
     regarded as an accurate, but partial,    way in which the campaign is being          tone for each actor, in which a score of

80   OPTIONS POLITIQUES
     NOVEMBRE 2008
Back to the future

zero reflects neutral coverage, positive                                          printed
                                             usual, a large proportion of the articles         about      this   campaign.
scores indicate increasingly positive        printed during this election were    Moreover, the prominence of the envi-
coverage, and negative scores indicate       about strategy and the horse race    ronment issue declined, slowly but
increasingly negative coverage. In these                                          steadily, as the campaign progressed.
                                             between parties and leaders. It is thus
data, the resulting measure ranges from      especially striking that the economy      Foreign affairs, crime and arts fund-
roughly -30 to +30, reflecting articles in   figured so prominently in the supply ing were the next most salient issue
which as many as 30 percent of the                                                domains in the press, appearing in
                                             of election news about this campaign.)
                                                                                              about 18 percent, 12 percent
The “Green Shift” policy was the centrepiece of the Liberal                                   and 10 percent of election
Party campaign. Moreover, as we shall see, the Green Party                                    stories respectively. Much of
figured more prominently than ever in news media. Despite                                     the foreign affairs coverage

these factors, environmental issues appeared in less than 2 of concerned                                    Canada’s ongo-
                                                                                              ing military commitment
every 10 articles printed about this campaign. Moreover, the                                  (and the cost associated) to
prominence of the environment issue declined, slowly but                                      the Afghanistan mission.
steadily, as the campaign progressed.                                                         Crime policy was highlight-
                                                                                              ed by coverage of the
words in sentences about a given party         The environment was the second     Conservatives’ proposed changes to the
or leader are negative or positive.        most salient topic. Even appearing in  Youth Criminal Justice Act, and arts fund-
Because the vast majority of articles are  nearly 20 percent of articles over the ing attracted media attention in
neutral, however, a weekly average for     campaign (including both environ-      response to Conservative campaign
net tone reflects a much narrower          ment and energy), however, it trailed  announcements.
range. Trends are apparent even using      the economy by a rather wide margin.        The shifting issue focus of the
that measure, though for the sake of       This is surprising given the pre-elec- campaign is illustrated in figure 1,
simplicity we use data here in which       tion salience of the issue, and given  which tracks weekly coverage for “all
articles are assigned to one of three cat- that the “Green Shift” policy was the  environment” stories (combining
egories: negative (“net tone” of less      centrepiece of the Liberal Party cam-  environment and energy), “all for-
than -2), positive (“net tone” of more     paign. Moreover, as we shall see, the  eign affairs” (combining foreign
than +2) and neutral (all others).         Green Party figured more prominently   affairs and international trade), and
     Note that given the expansive         than ever in news media. Despite these “all economy” (combining economy,
nature of the tone dictionary, this        factors, environmental issues appeared employment and financecommerce).
measure reflects a combination of          in fewer than 2 of every 10 articles   The figure shows not just the slow
content about the issues and events
being reported on, and the opinions
or attitudes expressed about the con-      FIGURE 1. TOPICS BY WEEK
tent itself. To be clear: party or leader                             80
tone can result from attention to
campaign gaffes and policy failures,                                  70
or from criticism of campaign plat-
forms. Thus, tone of coverage is a                                    60
summary measure of the relative neg-
                                             % of election articles

ativity of the events or issues being                                 50

covered and the opinions and atti-
                                                                      40
tudes of newsmakers.
     That is the 2008 study, then; what
                                                                      30
were the results? First, press coverage
during this campaign was dominated
                                                                      20
by economic issues. The economy fig-
ured in more than one-fourth of news                                  10
reports about this campaign. Indeed, if
we combine mentions of finance,                                        0
commerce and employment in this                                            Week 1    Week 2       Week 3           Week 4    Week 5            Week 6

category, economic issues were fea-                                                 All economy       All foreign affairs   All environment
tured in about 50 percent of election-
related news articles. (Note that, as        Source: The Media Observatory, McGill Institute for the Study of Canada.

                                                                                                                                      POLICY OPTIONS          81
                                                                                                                                              NOVEMBER 2008
Blake Andrew, Lori Young and Stuart Soroka

     decline in environment stories, but         Dion did not dominate the news                               tion by mainstream newspapers
     the rather striking prominence of the       agenda in 2008. The declining promi-                         occurred in spite of his position as
     economy, which steadily assumed             nence of the Liberal Party’s leader is                       leader of the opposition and of the
     control of the media agenda as this         quite remarkable. In 2004, almost                            second-largest party in the country.
     campaign unfolded. During the week          half of the news stories in the                                   Most of the articles concerning
     before the official kickoff at Rideau       Canadian press about leaders men-                            party leaders were led by coverage of
     Hall, economic issues already figured       tioned the Liberal Party leader first.                       Conservative leader Stephen Harper. In
     in roughly 4 in 10 news reports relat-      In 2008, Dion was the first leader                           fact, articles that began with Harper out-
     ed to the looming election. But by          mentioned in articles only about 15                          numbered all articles beginning with the
     the final week of the campaign, the         percent of the time. This marginaliza-                       other party leaders combined. Where
     economy was clearly at the forefront
     of the press’s attention. Indeed, in        FIGURE 2. LEADER FIRST MENTIONS
     the week leading up to voting day on
     October 14, economic issues were                                               50

     mentioned in more than 70 percent                                              45
     of articles about the campaign.
                                                                                    40

     T     here can be little doubt that the
           increasing coverage of the
                                                                                    35
                                                 % of election articles

     Canadian economy was directly related                                          30

     to concerns about the financial crisis in                                      25
     the United States that emerged during
     the campaign. In some ways, then, the                                          20

     2008 Canadian campaign looked a lot                                            15
     like the famous FTA campaign 20 years
     ago — the 1988 campaign was also                                               10

     dominated by questions about the                                               5
     economy, and especially the conse-
     quences of the Free Trade Agreement                                            0
                                                                                         2004                 2006                      2008
     with the United States. There are some
     fairly stark differences between 1988                                                      Martin/Dion          Harper    Layton

     and 2008, however (and not just that
                                                 Source: The Media Observatory, McGill Institute for the Study of Canada.
     the ’88 election was about trade rather
     than the domestic economy). For
     instance, in 1988 the main parties and      FIGURE 3. PARTY FIRST MENTIONS
     leaders all anticipated and primed the
                                                                                    45
     economy frame. It is not clear that the
     economy’s presence as the critical ballot                                      40
     box question in 2008 was either antici-
     pated or welcomed by any of the parties                                        35

     or leaders contesting this campaign.                                           30
                                                           % of election articles

          Another unique element of 2008
     campaign coverage is the salience of                                           25
     parties and leaders. To capture this, we
                                                                                    20
     rely on “first mentions” — the party or
     leader mentioned first in each article.                                        15
     First mentions indicate who is framing
     the election campaign — whether the                                            10
     article is about Harper and reactions to
                                                                                     5
     Harper, for instance, or, alternatively,
     about Dion and reactions to Dion.                                               0
     Figures 2 and 3 show that 2008 was                                                  2004                 2006                      2008

     clearly about the former.                                                                   Liberal       Conservative    NDP
          Unlike his predecessor Paul
     Martin, Liberal Party leader Stéphane       Source: The Media Observatory, McGill Institute for the Study of Canada.

82   OPTIONS POLITIQUES
     NOVEMBRE 2008
Back to the future

headlines were concerned, the gap was       prominently in the press than the                                                      stories during the previous two feder-
even larger. That is, for Canadians skim-   NDP. This was clearly a breakthrough                                                   al campaigns.
ming headlines, in both local and           year in terms of press coverage for
national newspapers, Harper was even
more predominant. Strictly in terms of
the proportion of coverage, Harper was
                                            the Green Party and its leader
                                            Elizabeth May. Overall, almost 5 per-
                                            cent of press articles during the cam-
                                                                                                                                   H     ow were these changes in the
                                                                                                                                         prominence of parties and lead-
                                                                                                                                   ers in 2008 related to the tone of cov-
mentioned three times as often as Dion      paign     mentioned       May     first,                                               erage? After all, prominence in the
by the end of the campaign.                 remarkable considering that her                                                        press does not necessarily mean posi-
     Party salience was also consider-      predecessor Jim Harris was men-                                                        tive press coverage. In 2004 and
ably different in 2008, although the        tioned first in only a tiny fraction of                                                2006, for instance, Paul Martin was
shift was not as dramatic as it was for
leaders. Once again, though, note in
                                            FIGURE 4. NEGATIVE COVERAGE OF LEADERS
figure 3 the consistent decline in the
Liberal Party’s share of the media                                                        25
spotlight during the three most
recent campaigns. In 2004, more                                                           20
than 40 percent of articles that
                                                            % positive minus % negative

involved a party mentioned the                                                            15

Liberals first. By 2008 the Liberals
                                                                                          10
accounted for only one-fourth of first
parties mentioned. They have
                                                                                           5
unquestionably been supplanted by
the Conservatives as the most salient                                                      0
party in the campaign. Some of this
shift may be a function of incumben-                                                       -5
cy, of course — it may be that the
party of government always has an                                                         -10
advantage where campaign framing
is concerned. Note, however, that the                                                     -15
                                                                                                 Week 1    Week 2         Week 3             Week 4            Week 5           Week 6
salience of the Conservatives has
been relatively consistent since 2004,                                                                        Harper              Dion               Layton              May

hovering around 30 percent of first
                                            Source: The Media Observatory, McGill Institute for the Study of Canada.
party mentions. While the promi-
nence of their leader, Stephen Harper,
has     risen      considerably,      the   FIGURE 5. NEGATIVE COVERAGE OF PARTIES
Conservative Party brand has not
                                                                                          15
become more salient in press reports
since 2004. The shift in party stand-
ings is due not to a rise in                                                              10
                                              % positive minus % negative

Conservative mentions, but to a
decline in Liberal ones.
                                                                                           5
     NDP and Green Party leaders
were also treated differently in terms
of prominence in 2008. Jack Layton’s                                                       0
salience in newspapers has declined
over time. His zenith appears to be in
                                                                                          -5
2004, when he was mentioned first
in about 12 percent of articles. The
2008 coverage of the NDP shows that                                                 -10
Layton was much less prominent in
the press. Only about 6 percent of
                                                                                    -15
articles led off with reference to the                                                          Week 1    Week 2         Week 3             Week 4            Week 5            Week 6
NDP leader specifically. Indeed, in                                                                       Conservative            Liberal             NDP               Green
the first few weeks of the campaign
the Green Party was featured more           Source: The Media Observatory, McGill Institute for the Study of Canada.

                                                                                                                                                                       POLICY OPTIONS       83
                                                                                                                                                                            NOVEMBER 2008
Blake Andrew, Lori Young and Stuart Soroka

     the most prominent leader, yet he          with net tone for Bloc Québécois leader  part of the campaign, however. That
     was also the leader who received the       Gilles Duceppe, since we are looking at  the Conservatives were unable to
     most negative coverage. Stephen            English-language papers only. What       reopen a gap between themselves and
     Harper received much better coverage       data we do have for Duceppe thus do      the Liberals as the campaign pro-
     in those elections, though he              not reflect his coverage for the vast    gressed proved to be rather important.
     appeared in far fewer articles than        majority of the BQ constituency.)        To be sure, the volume of Conservative
     Martin did.                                     Differences across parties are rel- coverage was extraordinary. But the
         A key difference in 2008 relative      atively similar. Conservative net tone   tone was, relative to their main com-
     to previous campaigns is that the          over the campaign was 4.7 points         petitor, rather middling.
     Conservatives and their leader             below the cross-party average, while          What do these results tell us
     received rather more negative cover-       for the Liberals it was -4.9; for the    about the 2008 election? They cer-
     age than they have been accustomed         NDP it was above 0.1 the average.        tainly highlight the increasing signif-
                                                                                                      icance of the economy
     Most of the articles concerning party leaders were led by                                        over the campaign. To the
     coverage of Conservative leader Stephen Harper. In fact,                                         extent that the campaign
                                                                                                      provides some kind of
     articles that began with Harper outnumbered all articles                                         mandate, it is clear that the
     beginning with the other party leaders combined. Where                                           next government is sup-
     headlines were concerned, the gap was even larger. That is,                                      posed to deal with the
     for Canadians skimming headlines, in both local and national economy. Results also illus-
                                                                                                      trate what appears to be a
     newspapers, Harper was even more predominant. Strictly in                                        failure on the part of the
     terms of the proportion of coverage, Harper was mentioned                                        Conservatives — a failure,
     three times as often as Dion by the end of the campaign.                                         that is, to fully capitalize
                                                                                                      on the sheer volume of
     to. Figures 4 and 5 illustrate trends in   Once again, the Green Party benefit-     Conservative Party and especially
     tone. Both figures rely on a relatively    ed most from newspaper coverage.         leader coverage. This was clearly a
     simple version of average net tone: the    The overall net tone for the Green       campaign in which media coverage
     percentage of positive stories minus       Party was 9.7 points above the aver-     focused on the Conservatives. This
     the percentage of negative stories for     age. In short, if a story mentioned      was not driven by what appears to be
     each party and leader. But to focus on     the Greens, or Elizabeth May,            overly critical media — certainly, cov-
     inter-party differences (and also filter   chances were high that coverage was      erage was no more critical than it was
     out the effects of shifting issue cover-   framed in positive terms.                for the Liberals. But at no point once
     age on our automated tone measure),            The most interesting trend over      the campaign was under way were the
     figures 4 and 5 show the weekly differ-    the campaign is the absence of any       Conservatives or Prime Minister
     ence between party/leader tone and         considerable gap between the             Harper able to garner markedly better
     the average of election tone for that      Conservative and Liberal parties.        coverage than the Liberals or Dion.
     week’s coverage of all main parties or     Indeed, in the final week of this cam-   The consequence was, perhaps, that
     leaders combined. The interpretation       paign the Conservative Party and         while the Liberals clearly did not win
     is as follows: a Liberal Party value of    Stephen Harper received the worst        the election, the degree to which the
     -12 in week 1 reflects the fact that       coverage in the major newspapers.        Conservatives did was rather limited.
     Liberal net tone was 12 points worse       But in spite of the Liberals’ compara-
     than the party average for that week.      tively rough start, from week 2          Blake Andrew is a doctoral candidate in
     Note that these automated results are      onwards there is rarely much notice-     political science and research coordinator
     not directly comparable with our           able difference in the tone of cover-    and Ph.D. fellow at the Media
     manual data from past elections; they      age for Liberals and Conservatives, or   Observatory, McGill University. Lori
     nonetheless provide a reliable indica-     for Dion and Harper.                     Young is a former graduate student in
     tion of trends, both over time and             This trend in media coverage nice-   political science and fellow at the Media
     across parties, in the 2008 campaign.      ly captures what voters observed in      Observatory, now working as a consult-
          Overall, net tone was as negative for both opinion polls and election com-     ant on media analysis. Stuart Soroka is
     Harper as it was for Dion. Combining all   mentary over the campaign. The           associate professor and William Dawson
     six weeks, Harper’s net tone was 6.3       Conservatives began with an advan-       Scholar in Political Science at McGill
     points below the average, compared         tage over the Liberals in terms of posi- University. Along with Antonia Maioni,
     with -6.4 for Dion, +2.3 for Layton and    tive versus negative coverage. This      he is co-director of the Media
     +10.3 for May. (We do not deal here        advantage faded quickly in the first     Observatory.

84   OPTIONS POLITIQUES
     NOVEMBRE 2008
You can also read