Effects of General Presidential Campaigns on Voters , 1980 - 2000

We used National Election Studies (NES) data from the six most recent presidential campaigns to ascertain the effect of campaigns on issue knowledge, evaluations of the Republican and Democratic candidate, issue salience (the number of issues used to evaluate the two candidates) and character salience (the number of character traits used to evaluate the two candidates). The analysis correlated the days passed in the campaign with levels of the dependent variables. Results show that campaigns are capable of, but do not necessarily, affect each of these variables. Furthermore, Republicans, Democrats, and Independents usually exhibited different effects. Effects related to policy (issue learning, policy salience) occur more frequently in voters than effects related to character (character evaluation, character salience). Implications for campaigns, and the study of campaign effects, are discussed. Key Terms: Campaign effects, issue knowledge, character evaluation, issue salience, character salience, political party Glenn Leshner, William Benoit and Gleen J. Hansen 47 Introduction The quadrennial selection of the president of the United States is an extremely important event in the life of our democracy. Although presidential candidates at times adopt similar policy positions (who would oppose protecting Social Security?), there are differences between them. In 1996, for instance, President Bill Clinton favored selective tax cuts whereas Senator Bob Dole advocated a 15% across-the-board tax cut. In the 2000 contest, George Bush recommended privatizing (at least some of) Social Security, a proposal Al Gore rejected. These examples highlight the fact that it can make a difference who is elected president. This raises the question of how our citizens learn about the candidates for the Oval Office and their policy positions. In the most recent campaign, the Democratic nominee was Al Gore, and Vice Presidents are simply not as well known as their running mates. Governor George W. Bush was not very well known outside of Texas. Relatively few voters can be considered “political junkies” who actively search for information about the candidates; most voters learn about the two candidates for president from whatever information they happen to encounter before election day. In contemporary society, virtually no citizen learns about the candidates’ policy proposals, or forms impressions about the candidates’ character, from direct, face-to-face contact with candidates. Political knowledge and attitudes, therefore, arise from media campaigns. Figure 1 depicts the flow of information in a presidential campaign. Candidate messages and the news media are the primary sources of information (and are depicted on the left; information tends to flow from left to right). Some of what the candidates say is filtered or interpreted by the media (arrow 1). Candidate messages (arrow 2) and news messages (arrow 3) at times reach voters directly (individual voters, of course, are exposed to different amounts and combinations of these messages). Both candidate (arrow 4) and news (arrow 5) messages are filtered and interpreted by other people, ultimately reaching voters through political discussion (arrow 6). 1 Notice that there are multiple forms of candidate messages (e.g., television spots, debates) and of news messages (e.g., television news, newspapers). There are also multiple instances of each (usually more than one debate, different television spots, daily newscasts or newspapers with different content). Each voter encounters an individual amalgamation of campaign messages. Of course, the modern presidential campaign is an intense (and lengthening) period of messages from candidates, news media, and others. Early scholars advocated the “limited effects” model of the mass media in general, and presidential campaigns specifically (see, e.g., Campbell, et al., 1960; Klapper, 1960; Lazarsfeld, Berelson, & Gaudet, 1944; but cf. Holbrook, 1996; Zaller, 1996). More recently, Lichtman (1996) declared, “Despite the hundreds of millions of dollars and months of media attention lavished on them, general-election campaigns don’t count” (p. 5). We take the contrary position on this issue, offering evidence that general election campaigns are capable of influencing voters. 1 This diagram does simplify matters in places. For example, it omits messages from third parties like special interest groups (but of course those messages are mediated as well). Similarly, occasionally some information will flow in other directions (e.g., candidates can make use of, or respond to, information initiated by news media; both candidates and media solicit information from voters, particularly using polls). Nevertheless, we believe it is a useful model of how information reaches voters. Presidential Campaigns 48 Specifically, we will employ American National Election Studies (NES) data, a national sample of adults, from the last six presidential campaigns to investigate the effects of campaigns on voters. Other studies have used NES data to examine the effect of particular sources of information on voters, like newspapers, television news, television spots, debates, or political discussion (e.g., Brians & Wattenberg, 1996; Holbert, Benoit, Hansen, & Wen, 2003). However, rather than try to identify the specific source of information, we seek only to discover changes in voters that occur during the course of the campaign. These data sets are valuable because they represent national samples of adults and may be considered to be a set of six replications. The combined N for the six campaigns we studied is 11,909. Theoretical Underpinning The Functional Theory of Political Campaign Discourse (e.g., Benoit, 1999; Benoit, Blaney, & Pier, 1998) argues that political campaigns help citizens decide how to vote by providing them with information about two topics: policy and character. Benoit and Harthcock (1999) posit that policy concerns “governmental action (past, current, or future) and problems amenable to governmental action,” while character concerns “characteristics, traits, abilities, or attributes of the candidates” (p. 346). Of course, there can be overlap between these two concepts (see Hacker, Zakahi, Giles, & McQuitty, 2000), but the concepts are distinct conceptually and content analysis reliably distinguishes campaign messages discussing the two topics (e.g., Benoit, 1999). Accordingly, this study will investigate possible effects of presidential campaigns on policy, or issue knowledge, and character, or evaluations of the candidates. We also investigate issue salience and character salience, which are measures of the number of issues or character traits used by voters to evaluate the candidates. Furthermore, we are interested in the question of whether campaign effects are the same for different groups of voters (Republicans, Democrats, Independents). We answer six specific research questions: RQ1. Does issue knowledge of the Republican candidate increase during the general election campaign? RQ2. Does issue knowledge of the Democratic candidate increase during the general election campaign? RQ3. Does evaluation of the Republican candidate’s character increase during the Glenn Leshner, William Benoit and Gleen J. Hansen 49 general election campaign? RQ4. Does evaluation of the Democratic candidate’s character increase during the general election campaign? RQ5. Does issue salience increase during the general election campaign? RQ6. Does character salience increase during the general election campaign? These questions will be answered using NES data from the 1980-2000 presidential campaigns. Method Our method will be to correlate the days passed in the campaign with the level of these variables (more precisely, the number of days that have passed from when the NES began collecting data in that campaign). If campaigns increase knowledge (or candidate evaluations, or salience), the correlation will be positive. If there is no change in these variables as the campaign progresses, the correlation will not be significant (and negative correlations would mean that the campaign reduced knowledge, hurt evaluations of candidate character, or reduced salience). Of course, this approach cannot identify the source of changes in these dependent variables. If knowledge of the candidates’ issue position increases, that learning could come from news coverage, candidate speeches, debates, political advertisements, political discussion among voters, or other sources. However, our purpose in this study is not to ascertain whence changes (if any) arise, but whether changes occur in the campaign. We believe that if significant changes in the levels of these variables occur over time, some source or combination of sources must be responsible. The question we take up is whether campaigns affect voters, not how or through which media. Dependent Variables The following dependent variables were measured in the pre-election phase for each NES general campaign survey. Issue knowledge: Issue knowledge was the sum of items that asked respondents to place each candidate on a scale (most often from 1 to 7) to demonstrate the level of support they believe the candidate had on a particular issue. We used only those issues where the two major party candidates clearly differed in their positions stated during the campaign. The number of items used for each level of issue knowledge varied across data sets, ranging from four items in 1992 to nine items in 1996 (see Table 1). Responses that indicated a correct identification of a candidate’s stand on issues were given one point. “Correct” answers were defined as reporting any level of agreement with an issue for which the candidate is in favor or any level of disagreement with an issue for which the candidate is not in favor (for 7-point scales correct responses ranged from 1 to 3 or from 5 to 7, depending on the issue and the candidate). “Incorrect” items were given a 0. Presidential Campaigns 50 Table 1: Issue Knowledge Items by Election Campaign 1980 defense spending 1992 defense spending government services government services/spending inflation/unemployment abortion abortion job assurance income taxes 1984 governmental services/spending 1996 defense spending minority aid government services/spending involvement in Central America health insurance defense spending job assurance social/economic status of women social/economic status of blacks cooperation with Russian crime standard of living environment environmental regulation abortion 1988 governmental services/spending 2000 government services/spending defense spending defense spending government-funded insurance job assurance standard of living social/economic status of blacks social/economic status of blacks abortion social/economic status of minorities environment cooperation with Russia gun control environmental regulation Character evaluation: Character evaluation was an index computed as the mean of responses to items that described character traits for each major party candidate. The number of items used for each level of character evaluation varied across data sets, ranging from six items in 1996 to 16 items in 1984, and included such characteristics as intelligent, compassionate, moral, weak, power-hungry, etc. Items were measured on 4-point response scales anchored by “extremely well” to “not well at all” and were recoded where necessary such that a higher value represents a more positive character evaluation. The alpha for character evaluation scales ranged from .79 (Carter in 1980) to .96 (Reagan in 1984). See Table 2. Issue and character salience: Each respondent was asked if they “like anything about (candidate’s name),” for each candidate. They were also asked if they “dislike anything about (candidate’s name),” for each candidate. If the respondent replied “yes,” they were probed for details (up to 5 mentions for each candidate). Each detailed answer was coded such that mentions about issues were coded differently than mentions about character. Each issue mention was given a 1, then all mentions were summed across candidates for a total issue salience score. The same was done for character salience. Grouping Variable Party identification: NES asks respondents a series of three questions to determine party identification (one ID question, then two follow-ups to determine strength). From those Glenn Leshner, William Benoit and Gleen J. Hansen 51 items, NES creates a summary party ID item (range = 0-6, where 0 is strong Republican and 6 is strong Democrat) that formed the basis of our analyses. Zero and one were computed as Republicans, five to six were computed as Democrats, and three to five were computed as Independents. Table 2: Character Evaluation Items by Election Campaign 1980 Moral 1992 Intelligent Dishonest Compassionate Weak Moral Knowledgeable Inspiring Power-hunger Strong leadership Inspiring Cares about people Strong leadership Knowledgeable Republican ∝ = .82 Honest Democrat ∝ = .79 Gets things done Republican ∝ = .9