«13TH INTERNATIONAL PUBLIC RELATIONS RESEARCH CONFERENCE “Ethical Issues for Public Relations Practice in a Multicultural World” Holiday Inn ...»
Methods To answer the questions above, the study examined issue mentions in the speeches and political advertisements used by the Barack Obama and John McCain campaigns during the 2008 presidential election. Using issue categories revised from the codebooks for the National Election Studies data from the University of Michigan (Miller, Kinder, Rosenstone, & the National Election Studies, 1993), campaign speeches and political advertisements were content analyzed for issue and character mentions. The 46 categories consisted of issue keywords and phrases for three general areas: domestic issues, foreign policy issues and those focusing on personal character. The codebook has been used in previous elections (Boyle, 1998) and revised to include issues relevant to the 2008 campaign (e.g. “war on terror”). The campaign materials were divided into an early campaign season time period and late campaign season time period so intracandidate (within each campaign) and intercandidate (between campaigns) agenda-setting could be examined. Both campaign periods covered the traditional campaign season beginning officially after Labor Day. The early campaign period (time one) covered speeches and advertisements from September 3 to October 3, 2008. The later period (time two) covered October 4 through election day (November 4, 2008).
The advertisements were transcripts created from the narratives of the 2008 presidential advertisement collection from the Political Communication Lab at Stanford University website (Political Communication Lab, 2009). The publicly available website archives all of the 2008 general election political advertisements by date (http://pcl.stanford.edu). Transcripts for the speeches were downloaded from the Lexis-Nexis database using the time period noted above and general descriptors (e.g. Obama) from Congressional Quarterly (CQ) transcripts. Since both major presidential candidates were sitting U.S. senators, speeches from their campaign events and U.S. Senate floor remarks were available. The total time one advertisements was 58 and an additional 41 in the later time period. The word total for the Obama ads was 4,795 (time one=2,646; time two=2,149) and McCain was 4,175 (time one=2,345; time two=1,830). Most of the political advertisements were 30-second spots so their totals were much less than those of the campaign speeches. There was a total of 53 time one speeches and 96 in time two. Word counts for the speeches were 192,215 (time one=64,763; time two=129,452) for McCain and 238,787 83 for Obama (time one=97,479; time two=141,308). Given the extensive nature of the content analyzed, Concordance software was used to search for the extensive keywords and key phrases listed in the codebook (see http://www.concordancesoftware.co.uk). Similar content analysis software was used in previous studies considering intercandidate agenda setting (e.g. Dunn, 2009, Tedesco, 2005). Communication researchers have considered the effectiveness of computer coding software and how it compares to traditional content analysis coding (e.g.
Nacos, Shapiro, Young, Fan, Kjellstrand & McCaa, 2009). Computer software allows the analysis of data quickly, but has limitations of not being able to consider contextual factors (Nacos, et al., 2009, pp. 251-252). Krippendorff has noted the value of combining both kinds of analysis and calls this “computer-aided textual analysis, or CATA for short.” (2004, p. 261). The analysis conducted here used the Concordance software to identify the frequencies of keywords and related phrases with the researcher reviewing them in their context. Great care was taken to develop content categories that did not overlap. For example, the word “bridge” is too broad since it could relate to the category Infrastructure and also Leadership when a candidate indicated wanting to build a bridge to the future. After extensive revisions based on pilot study analyzes, the codebook was used to consider 46 (later reduced to 43) categories of issues. These methods led to high levels of face validity. A second coder used Concordance software to content analyze more than 20 percent of the data and intercoder reliability was nearly 100 percent.
After the category mentions were totaled for the speeches and advertisements, the data was entered into a statistical software data set. The results presented in the paper are shown descriptively through frequencies and inferentially through correlation coefficients. Along with considering traditional statistical significance for the Pearson product correlations, the analysis considers the influence across time. Tipton, Haney and Baseheart (1975) and many scholars have utilized the Rozelle-Campbell test for statistical significance. 3 Weaver, Graber, McCombs, and
Eyal (1981, p. 136) explain its function:
The general procedure for calculating the baseline is to average the synchronous (same-time) correlations between the measures at each of the two points in time and to average the autocorrelations (over-time correlations) between the same measure over time. This provides a baseline estimate of the diagonal correlations that one should expect to observe if there was no causal relationship between one variable at Time 1 and changes in the other between Time 1 and Time 2.
The use of this statistical technique allows the study to better answer the research questions considering their cross-lag nature.
Results Tables one through four begin to address the first research questions considering intracampaign agenda setting. Considering table one, McCain’s speech issue priorities did show some
change between early and late in the campaign. The economy and taxes switched places on the lists with taxes becoming more of a focus later in the campaign. In addition, healthcare becomes a greater issue later in the campaign moving up from seventh in time one. Table two indicates taxes were the top priority of the narratives of political advertisements for McCain 2008. His leadership abilities were stressed in the advertisements more than in the speeches.
Tables three and four consider the focus of Obama’s campaign speeches and advertisements in 2008. Like McCain, taxes were a major issue in his speeches and advertisements. In his speeches, Obama focused on it more than any other issue both early – unlike McCain – and late in the campaign. In time one of the speeches, Obama did not mention the campaign but focused more on it as election day came closer and pushed education as a topic from the top five. In his paid messages, the economy and taxes was of great focus with leadership, employment and health remaining throughout both periods.
N=43. Unit of observations are public speeches and political advertisements. Including both campaigns, total Time 1 speeches = 53; Time 2 speeches = 96; Time 1 advertisements = 58;
Time 2 advertisements = 41. Time 1= Sept. 3 – Oct. 3, 2008. Time 2 = Oct. 4 - Nov. 4, 2008.
All of the Pearson product-moment correlation coefficients described below were statistically significant at the p≤.01 level. To further explore the cross-lag influence of messages in time one and two, the Rozelle-Campbell baseline statistic was computed. If there were influences from time one, the coefficient (thick arrow in figures) would be expected to be larger than the baseline and larger than the reverse relationship (thin arrow). Figures one and two focused on research question one which asked if the messages within the individual campaigns (intracampaign) were consistent. As indicated by the large and statistically significant coefficients there were moderate to strong relationships between advertising messages and campaign speeches. Campaign speeches provide more of an opportunity to test messages and issue positions on the public and, because of this ability, the cross-lagged comparisons considered if the speeches in time one led to advertising messages in time two. For the McCain campaign, the baseline and expected relationship was statistically the same --.51 versus the baseline of.52. Given the high correlation of the candidate messages in time one (.85 and.81), partial correlation coefficients were computed holding constant the influence of the message in the earlier time period as appropriate for cross-lag analysis (e.g. Chang, 2007). In figure one, the expected relationship failed to reach statistical significance (p=.10) while the reverse relationship produced a large partial coefficient (p=.00). There is support for the influence of the McCain advertisement messages in time one on the McCain speech mentions in time two. In figure two, both of the partial correlations were statistically significant but the reverse relationship indicating the influence of the advertisement message on the later speech mentions was more than double the size of the expected relationship. Again, support exists for the influence of the advertising mentions in time one on the later speech mentions of both campaigns.
The correlation analysis found in figures three and four seek to answer research questions two and three by examining the intercandidate (between campaigns) agenda-setting. Of focus are messages found in advertisements and speeches for a campaign in time one on the other major candidate’s campaign messages in time two. In figure three, McCain’s speech mention expected correlation was substantively larger than the baseline statistic, but statistically the same as the reverse coefficient (.82 versus.81). Figure three indicates two significant, but equal relationships. McCain speech mentions at time one had the same influence as Obama’s speech mentions at time one on their opponent’s time two speech mentions. When controlling for within 86 campaign earlier time periods (.94 and.81), the partial coefficents are both significant. The.82 size of the one in the expected direction is larger than the.63 for the reverse relationship (both p=.00). There was greater support for the influence of McCain’s speech issue messages on Obama’s speech messages than the reverse.
Figure four considers the influence of McCain advertisement messages on Obama. The expected coefficient is substantively higher than the baseline statistic and the reverse coefficient.
Furthermore, the partial coefficient for the expected relationship was.56 and significant (p=.00) while the reverse relationship produced a coefficient much smaller and not significant. These results support the belief that McCain’s earlier advertising messages had an influence on Obama’s advertising messages in the later time period.
In the last pair of figures, intercandidate messages are considered across media type -advertisements versus speeches between the campaigns. Both the expected and reverse relationships are above the baseline statistic, but only marginally. Partial coefficients fail to reach statistical significant and are small. Figure six reflects a similar; both are slightly above the baseline and have similar Figures 1-6: Cross-Lagged Comparisons (Pearson Coefficients) 4 Figure 1, Intra-Campaign: McCain Speech versus Advertisement Mentions Rozelle-Campbell Baseline =.52
4Each line represents a Pearson coefficient. The lines in bold indicate the coefficient of primary interest and the other arrow line represents the area of interest if the reverse hypothesis was true. The RozelleCampbell statistic was computed for each model to determine the statistical significance of individual Pearson coefficients.
87 Figure 2, Intra-Campaign: Obama Speech versus Advertisement Mentions Rozelle-Campbell Baseline =.79
in strength. Again, partial coefficents fail to reach statistical significant, are small, and provide no support for intercandidate agenda-setting across media type.
The analysis here addressed several suggestions from previous researchers. The cross-lag correlations considered allowed consideration of differences between time – a limitation identified (Kiousis, Kim, McDevitt & Ostrowski, 2009). It also adds to the agenda-setting Figure 5, Intercandidate: McCain Speech versus Obama Advertisements Rozelle-Campbell Baseline =.75
scholarship by considering more than just one campaign public relations tool. Some previous studies used only one information subsidy to indicate the messages produced by campaigns and researchers noted this was a limitation of their analyses (Dunn, 2009; Tedesco, 2005). The use of political advertisements and speeches provide greater range to the campaign messages analyzed in 2008.
Within each campaign the advertising messages used in September influenced speech issue mentions in October. Given the abundance of polling data available to modern presidential candidates, it is possible campaigns felt their messages were having some effect and they reinforced them during their campaign speeches. Candidates often believe it is important to have a consistent message. Since advertising messages are generally delivered in 30-second lengths, they need to be a basic repetition of the core campaign communication. In this way, the findings in 2008 support the belief that advertising messages are “a parsimonious surrogate for the campaign as a whole” (Roberts & McCombs, 1994, p. 258).
The findings for the 2008 data support intercandidate agenda-setting within a media (subsidy) type for speeches and advertisements. This is different than the findings of Kiousis and Shield which did not find any intercandidate agenda-setting for speeches or position papers, but only for news releases (2008). In 2008, it was McCain whose mentions in time one influenced Obama’s messages in time two. The news media typically provide election coverage focusing on the “horse race” of the election (Hollihan, 2009, p. 130). Given this focus, the candidate trailing in the race often receives greater focus. The media coverage might provide an explanation why McCain’s speech and advertising messages in the first time period led to a greater frequency of issue mentions in time two by the Obama campaign.