Playing to the Crowd: Agenda Control in Presidential Debates

Abstract

Presidential debates allow candidates to send a message directly to voters. We use an experimental design complemented with a content analysis of all presidential debates in 1992, 2004, and 2008 to explore how candidates should and do use agenda setting, framing, and message tone to shape the agenda in debates. We find that candidates are differentially attentive to various topics, depending on the comparative advantage provided by the topic. Yet, this agenda control occurs only at the margins because topic salience in public opinion predicts candidate attention and conditions voters' receptiveness to debate rhetoric. Our findings thus suggest that topic salience constrains candidates' abilities to focus the agenda strategically.

[Supplementary material is available for this article. Go to the publisher's online edition of Political Communication for the following free supplemental resources: Experimental Sample and Randomization; Experimental Treatments; Experimental Results; and the Presidential Debate Rhetoric Codebook.]

Keywords:

• debates
• agenda setting
• framing
• tone
• salience

Notes

1. We can only go so far in inferring underlying strategy from observed messaging behaviors (and resulting rhetorical patterns). Still, the systematic rhetorical patterns that we observe can provide suggestive evidence of strategies candidates may have pursued and can identify strategies that candidates did not pursue successfully. Whether intentional or not, candidates' debate behavior sends important cues to citizens.

2. Riker derives this strategy from two general principles. First, his dominance principle asserts that candidates avoid topics on which the other side has an advantage. Second, the dispersion principle argues that both sides ignore topics on which neither holds an advantage (Riker, 1996). These principles conflict with the strategy of playing to the crowd because they suggest little issue convergence between opposing campaigns, whereas the playing-to-the-crowd strategy suggests competing campaigns will each address the same salient issues (Sides, 2006,p. 412).

3. We are able, however, to control for the partisanship and ideology of subjects, which is important because these fundamental political attitudes should condition subjects' receptivity to various topics and frames (Jost, 2006; Lakoff, 2006).

4. As coded by the Policy Agendas Project (www.policyagendas.org), based on raw Gallup survey results archived in the Roper Center iPOLL databank.

5. In the case where statements were vague or did not contain any substantive information, we coded the statement in the context of the surrounding discussion by reading the statements the candidate made both before and after the given statement. If the statement in question was clearly an extension of remarks that the candidate made before and/or after, we then coded the statement so that it was consistent with the overall message the candidate was conveying.

6. Content analysis of the 1992, 2004, and 2008 debates was conducted by four trained coders: Two coders completed the 2008 debates, and then one of these original coders as well as two additional coders completed the 1992 and 2004 debates. At least 300 of the statements coded by each coder were also coded by another coder, without the coders knowing which statements were being cross-coded. Pairwise tests showed strong intercoder reliability. Specifically, the minimum percentage agreements between coders on the variables of topic, frame, and tone were 94.6%, 85.1%, and 86.5%, respectively. The minimum Cohen's kappa scores for topic, frame, and tone were 0.922, 0.794, and 0.769, respectively. The minimum Krippendorff's alpha scores for topic, frame, and tone were 0.922, 0.795, and 0.768, respectively. These scores are based on a minimum of 75 statements in each pairwise test and include the full range of values for each variable. For topic and frame, an “other/not codeable” option was available; this code was employed for 4% of all candidate statements for topic and for 18% of all candidate statements for frame. A “neutral” option was employed for 28% of all candidate statements for tone; when cases of agreement about a neutral code are removed from the statements tested for intercoder reliability, coders demonstrate 85.5% agreement, a Cohen's kappa score of 0.73, and a Krippendorff's alpha score of 0.729.

7. The second debate in 1992, the second debate in 2004, and the second debate in 2008 all utilized the “town hall” format, with questions coming from the audience. In 2008, questions in this town hall debate were also posed via YouTube.

8. In the rare case that the candidate mentioned more than one policy topic in a single statement, we coded the statement according to the topic that dominated the statement. However, in the very few cases that the candidate gave two or more topics approximately equal consideration, we coded the statement according to the first topic mentioned.

9. After reading the debate text, subjects were asked to rate their overall support for each candidate using a 5-point scale ranging from 0 (very negative) to 4 (very positive).

10. This finding is based on additional analyses not presented in the main body of this article. These analyses also include: a model of the core results shown in Table 1; a replication of these results using an alternative dependent variable in the form of the number of like/dislike clicks subjects gave the second candidate while reading his response; and the regression analysis that produced the results for Figure 1.

11. In Figure 2 and all other findings presented, we consider only the two main candidates in each debate, excluding Perot's statements in 1992.

12. For example, an Ipsos-Public Affairs/McClatchy Poll conducted in October 2008 showed that 53% of respondents believed McCain was “engaging in more negative campaigning,” as compared with 30% who cited Obama as being more negative (see http://www.ropercenter.uconn.edu/data_access/ipoll/ipoll.html). A George Washington University Battleground survey conducted the same month showed that 57% of respondents believed that McCain was running a “somewhat” or “strongly” more negative campaign, with 20% citing Obama as being more negative.