Who Says What? Source Credibility as a Mediator of Campaign Advertising*

Shanto Iyengar and Nicholas A. Valentino

University of California, Los Angeles and University of Michigan


*We acknowledge the generous support of the National Science Foundation, the John and Mary R. Markle Foundation, and the University of California.

As every election day approaches, voters all across the country are forced to endure "saturation bombing" in the form of political advertisements from every manner of candidate and cause. During the 1996 campaign, for instance, the major presidential candidates aired 5,700 advertisements in the Los Angeles media market. Campaign advertising is now such a strong force that political analysts habitually attribute electoral outcomes directly to some facet of the candidates' advertising tactics. In 1996, leading newspapers diagnosed the "winning" strategy of the Clinton campaign's ad team the day after the election.

It is hardly surprising, of course, that the very political consultants who design the advertisements claim that these messages influence electoral outcomes. But more objective sources, including a substantial body of social science research, also support this conclusion. Whether one learns about it in the New York Times or the Public Opinion Quarterly, campaign advertising seems to matter.

Although there is an emerging consensus about the efficacy of advertising (see Shaw, 1997; Goldstein, 1997), little is known about the psychological mechanisms relating campaign advertisements to support for the sponsoring candidate. Most prior research has been preoccupied with refuting the "minimalist" conception of political campaigns (see Iyengar, 1996). Higher-order questions concerning the specific attributes of persuasive advertisements or the reasoning process employed by voters remain unanswered. This paper takes a first step toward understanding the mechanism of persuasion by specifying and testing one particular psychological account of the "winning" message.

Given its wealth of evidence on attitude change in a variety of contexts, the literature of social psychology is the natural springboard for political advertising research. This literature differentiates between message-based and cue-based accounts of persuasion. According to the former, the audience is prompted to consider the arguments contained in the message, retrieve relevant supportive or antithetical information from memory and adjust the "target" attitude appropriately. In cue-based models, the audience attends to more easily assimilated "signals" or associations that operate as decision making shortcuts. For example, cues can be derived from the gender or race of the sponsoring candidate; the spokesperson making the `pitch' for the candidate; attitudes concerning the sponsoring candidate's political targets; and vivid images, emotion-evoking music, humor, negativity and other nonsubstantive elements of a message.

The evidence from attitude change research indicates that in everyday situations involving efforts to persuade, cue-based processing predominates. Thus, in most situations, the recipient accepts or rejects persuasive messages based on an evaluation of the source's credibility (the research bearing on credibility is reviewed by McGuire, 1986; for an extension to political attitudes, see Lupia, 1996). Presumably, reliance on cues is more efficient for the recipient of the message than assembling and processing arguments; reliable sources provide a relevant and simple shortcut for deciphering the information value of the message. However, when recipients are highly involved in the persuasion situation (as, for instance, when the decision at hand represents a major financial commitment or when the recipient possesses high "expertise" on the subject) they typically focus more on the content or substance of the message than on associational cues. When the recipient is either motivated or expert, attitude change research finds that the logical or textual content of the message is more likely to influence attitude change than evaluations of the message's source (for reviews of this research, see Petty and Cacioppo, 1984; McGuire, 1985).

Given what we know about the typical voter's level of political engagement and familiarity with political appeals, the cue-based model of persuasion would seem especially applicable to the case of campaign advertising. Most voters lack the motivation and/or resources to acquire even the most elementary level of factual knowledge about the candidates and campaign issues (Popkin, 1991; Buchanan, 1991). They encounter information about the campaign not because they actively seek it out, but rather because it is sometimes difficult to avoid. In such low involvement environments, voters' reactions to visible cues -- including a candidate's party affiliation, gender or physical appearance -- take on special importance as mediators of campaign advertising (see McDermott, 1997). In terms of the classic shorthand of message learning theory -- "who says what to whom" -- "who" (source-related variables) can be expected to surpass "what"(message-related variables) as facilitators of political persuasion. The impact of political advertisements is more likely to be conditioned by what voters know, or are willing to infer, about the sponsor of the ads than the particular statements, claims or promises made in the ads.

Unlike the social psychological literature which typically treats source and message characteristics as unrelated, we argue that credible political advertising depends jointly upon who makes a statement and what is said. A credible advertisement is one that "resonates" with voters -- i.e., is congruent with their pre-existing beliefs about the competing candidates' strengths or weaknesses. Thus, in our view, campaign communication is most persuasive when it plays upon -- or interacts with -- voters' prevailing expectations or predispositions.

The most important political predisposition in the context of American campaigns is partisanship. Most Americans acquire a sense of party identification at an early age, and this psychological anchor remains with them over the entire life cycle (Jennings and Niemi, 1981; Sears, 1983). Not only are voters socialized to consider themselves Democrats or Republicans, they also learn to attribute differing policy positions to Democratic and Republican candidates. For instance, candidates are considered especially responsive to the interests of the groups or movements that comprise the party coalitions. Democrats are thought to represent the interests of blue collar workers, racial minorities, women, and environmentalists. Republicans are considered sensitive to the claims of businessmen, the wealthy, "born-again" Christians, and white males. Based upon these beliefs about responsiveness to group interests, voters also impute differences in the willingness and capability of the parties to resolve policy problems. Republicans are expected to exhibit a strong sense of fiscal responsibility, opposition to "big government," and a tough posture on crime and national defense. Conversely, Democrats are expected to reduce unemployment, maintain government social welfare programs, and help minorities and women overcome discrimination and inequality (Baumer and Gold, 1995; Petrocik, 1996; Ansolabehere and Iyengar, 1994; for an extension to European parties, see Budge and Fairlie, 1983; Kleinnijenhuis and de Ridder, 1997).

Such ingrained expectations about the commitment and performance records of candidates serve as important filters for interpreting and understanding campaign communication. In general, this argument suggests that political campaigns should be fitted to the party's reputation: a Democrat should be better off using advertisements that emphasize his intent to reduce unemployment, while a Republican should promote his support for a more punitive criminal justice system. Campaigns in which defense and military issues are central will work to the advantage of the Republican, while emphasis on social security and child-care will benefit the Democrat. In short, campaign communication that conforms to voters' expectations is most likely to be credible and compelling.

This paper assesses the role of credibility in political advertising by examining the ?fit? between the issue content of particular advertisements and the sponsoring candidate?s party affiliation. Our evidence stems from a series of field experiments carried out in southern California during the 1992 senatorial elections, the 1994 gubernatorial campaign, and the 1996 presidential campaign. The results consistently demonstrate that when candidates focus on issues that match their party affiliation, they enjoy significant electoral advantages.

In Study 1, we examined the electoral success of U.S. senatorial candidates when they broadcast a campaign advertisement that dealt with crime, unemployment or "women's issues". We found that campaign advertising on gender-related issues or unemployment was optimal for Democratic candidates Barbara Boxer and Dianne Feinstein, and that they gained least by advertising on crime. Our second study moved beyond the examination of a single candidate's advertising to focus on the campaign "dialogue" between the advertisements of gubernatorial candidates Kathleen Brown (the Democratic challenger) and Pete Wilson (the Republican incumbent). Here the results indicated that Ms. Brown fared best when both candidates converged on the issues of education or unemployment, and, conversely, that Wilson's prospects were brightest when the candidates advertised on crime or immigration. Finally, our third study examined the ability of presidential candidates Bill Clinton and Bob Dole to sway voters with advertisements that addressed the scope of government social welfare programs, taxes, or illegal drugs and immigration. In this case, the results showed that President Clinton's support peaked when the campaign agenda concerned government funding of social security and other benefit programs while Senator Dole was most effective when the candidates addressed issues of illegal drugs and immigration.


We rely on experimental methods. The advantages and disadvantages of experimentation are well known. Unlike surveys, experiments yield precise causal inferences about the effects of campaign advertising. The experimenter carefully selects the campaign messages that will be tested. Participants are assigned at random to conditions where they are exposed (or not exposed) to these specific campaign messages. This random assignment assures that the groups are equivalent in all respects but for exposure to particular campaign ads. These simple techniques provide researchers with the all-important ability to attribute any observed difference between the experimental and control groups to the effects of the experimental stimulus.

Of course, experiments have their own liabilities. Most are administered upon "captive" populations -- college students who must serve as guinea pigs in order to gain course credit. As the eminent experimental psychologist Carl Hovland warned many years ago (Hovland, 1959), college sophomores are not equivalent to "real people." A further weakness of the typical experiment is the somewhat sterile, laboratory-like environment in which it is administered. This research environment bears little resemblance to the cacophony and confusion of election campaigns.

We enhanced the realism and generalizability of our campaign experiments in several ways. Each of our experiments took place during an actual campaign and featured real candidates -- Democrats and Republicans, liberals and conservatives, males and females, incumbents and challengers -- as the advertising sponsors. Moreover, the advertisements that made up our experimental stimuli were highly realistic. They were selected either from actual advertisements used by the candidates during the campaign (in 1994 and 1996), or they were produced by us to emulate typical campaign advertisements (In 1992). In the case of our own productions, we spliced together footage from actual advertisements or news reports using studio-quality editing technology, making it difficult for all but the most sophisticated viewers to detect any differences between the experimental manipulations and the "real thing."

An especially important step toward boosting the generalizability of our results was the use of a subject pool that was reasonably representative of the Southern California voting-age population. Unlike the usual social science experiment, which relies heavily on conscripted college sophomores as subjects, our participants were people from many walks of life and included adults of all ages, employed and unemployed, whites, African-Americans, and Hispanics, men and women, city dwellers and suburbanites, etc.

The sites for each experiment were furnished to resemble, as closely as possible, the normal conditions in which a person views political advertisements. Comfortable couches and chairs were arranged in front of a television set, with houseplants and wall hangings placed around the room. Respondents were offered coffee, cookies, and soft drinks to enjoy during the viewing sessions. In most cases, family members or friends took part in the experiment at the same time, so that respondents did not find themselves sitting next to a stranger while viewing the political advertisements. Participants were recruited by the use of flyers, announcements in newsletters, and by personal contact in shopping malls offering payment of $15 for participation in "media research."

The sites selected for each experiment were virtually identical in layout and decor. Each site consisted of a two-room office suite located in or near a retail shopping area. One of the rooms was used as a viewing room and the other was used for filling out questionnaires.

In 1992, we administered the experiment in two locations. The first was located near Westwood, a predominantly Democratic neighborhood located just south of the UCLA campus. The other was located in Costa Mesa, a small city in conservative Orange County. In 1994, our site was located in downtown Los Angeles adjoining several major office complexes. In 1996, we expanded the facilities to include three sites. One was located again in Westwood, in a popular shopping mall. The second was based in a small shopping area in Moorpark, a conservative northern suburb of Los Angeles. The third site was located in Manhattan Beach, a coastal city southwest of Los Angeles. This variety of locations helped to ensure a large and diverse subject pool.

The Designs

We relied on two different experimental designs. The first (administered during the 1992 and 1994 campaigns) embedded the experimental advertisements into a fifteen-minute recording of a recent local newscast. Candidates advertise heavily during local news programs (because the audience for news includes a large proportion of likely voters) and the appearance of the experimental campaign advertisement in the local newscast was thus inconspicuous. Other news stories and product advertisements were selected carefully so that they were not relevant to the experimental stimulus, and these same filler stories and ads were used in combination with each political advertisement.

The "newscast" design was administered in two different versions containing either one or two campaign advertisements. In the one-advertisement design (employed in 1992), we limited the experimental manipulation to a single thirty-second commercial. In these studies we produced the experimental advertisement ourselves, the ad addressed either a "Democratic" or "Republican" issue, and we manipulated the resonance of the ad by airing it on behalf of both the candidates. One set of participants, for example, saw an advertisement aired by Democrat Dianne Feinstein that discussed rising unemployment and her support for job training programs and economic development tax incentives. Another set of participants saw the identical advertisement, this time on behalf of Feinstein's Republican opponent John Seymour. In other conditions, participants watched a Seymour (and Feinstein) advertisement that called for tough new anti-crime measures. In this way, we varied the degree to which one or other of the candidates' ads was especially credible.

The two-ad design (which was used in our study of the 1994 gubernatorial campaign) expanded the scope of the manipulation to include two advertisements, one from each of the candidates. Within this "paired" arrangement, we varied the issue agenda of the campaign, so that participants were exposed to spots dealing either with the Democratic issues of unemployment and education, or the Republican issues of crime and illegal immigration.

In 1996, we introduced a different one-ad design in which participants simply watched a videotaped collection of nine television advertisements one of which was an advertisement from the 1996 presidential campaign. The ads corresponded to those being aired in southern California by Bill Clinton and Bob Dole. The presidential ads we examined addressed illegal immigration, drug abuse, federal spending on social programs, and cutting taxes.


Our basic indicator of candidate preference was the respondent's intended vote choice following exposure to a campaign advertisement. This variable was coded as a trichotomy, with intended votes for Democratic candidates set equal to +1, votes for Republican opponents set equal to -1, and respondents who expressed no preference or who remained unsure of their vote being scored as zero. In effect, this indicator measures the Democratic candidate's lead over the Republican.

Since there were two concurrent Senate races in 1992, we pooled across both elections. That is, depending on the sponsor of the advertisement they were exposed to, respondents chose between Barbara Boxer and Bruce Herschensohn or between Dianne Feinstein and John Seymour. For respondents assigned to the control condition (in which there was no campaign advertisement), the vote choice indicator was defined as the average margin by which the Democratic candidates were favored in the two Senate races. As a result, the effect of the experimental manipulation in the 1992 studies is the difference between the lead of either Democrat in each advertising condition and the average lead of both Democrats in the control condition.

Democratic versus Republican Messages

We categorized advertisements as Democratic or Republican according to the reputations of the parties with regard to specific issues. As described earlier, we paired advertisements sponsored by each candidate in each election that concerned a specific issue. Our sample of Democratic issues included unemployment (1992 and 1994), education (1994), womens' rights (1992), and government funding for social welfare programs (1996). Unemployment was particularly salient in 1992 when the severe recession forced all four senatorial candidates to address the issue in their advertising. We created an ad in which the sponsoring candidate indicated his/her support for job training programs. A similar theme was prominent in the 1994 gubernatorial campaign as Brown attacked Wilson for the state's loss of jobs and Wilson claimed credit for having implemented various job-stimulus programs.

The year 1992 was hailed as the "year of the woman" in American politics. With two prominent women Democrats running for Senate seats in California, it was to be expected that gender-related issues would heat up as the campaign progressed. We produced an ad on behalf of both Democrats (but not on behalf of their Republican opponents) in which they stated their support for abortion rights and pledged to support legislation against sexual harassment in the workplace.

In 1994 we tested another traditionally Democratic appeal -- the quality of public education. Kathleen Brown aired ads criticizing the poor quality of California's public school system and promising to modernize teaching. Pete Wilson countered with ads pledging to protect state funding of public schools.

The final Democratic issue -- which featured prominently during the 1996 campaign -- focused on government spending for social welfare programs (most notably Social Security and Medicare). President Clinton repeatedly reminded voters of Republicans' efforts to weaken these popular benefits while Dole attempted to portray Clinton as a liberal Democrat who favored handouts and programs that created "big government."

Our sample of issues owned by Republicans included crime (1992, 1994 and 1996) and immigration (1994 and 1996). Of course, crime is a perennial issue in political campaigns and it was especially prominent during the 1992 senatorial campaigns in the aftermath of the Los Angeles riots. We created ads in 1992 that described the sponsor as a forceful advocate of "law and order." In 1994, Pete Wilson contrasted his support for "one strike" legislation for major crimes with his opponent's opposition to the death penalty. Kathleen Brown attempted to engage Wilson on the crime issue by airing an ad in which she blamed Wilson for increases in youth crime and described her proposal to build "boot camps" for juvenile offenders. A variation on the theme of crime took center stage in 1996 when Dole invested heavily in ads attacking the Clinton Administration's record on teen drug usage.

The issue of illegal immigration was used extensively during the 1994 gubernatorial campaign. Capitalizing on the publicity generated by Proposition 187, Wilson aired several advertisements citing his tough stance on the issue. Brown countered by attacking Wilson's proposal to deny the children of illegal aliens access to health care and education. Senator Dole revived the immigration issue in 1996 by airing ads (specifically targeted for California voters) that documented the significant increase in the number of illegal immigrants and the adverse consequences (including higher crime and overcrowding in the schools) of immigration for Californians.

In 1996, we also tested a number of ads from both candidates concerning taxation. Dole attempted to position himself as a tax cutter, while Clinton attacked Dole's tax cut proposal. The issue of taxes provides an interesting variant on the notion of resonance concerning temporary shifts in party ownership, especially with regard to economic issues. Traditionally, Republicans might be expected to own the tax issue. During the Reagan years, Republican candidates at every level of government successfully depicted themselves as "tax busters" while simultaneously painting their Democratic opponents as "tax and spend liberals." However, several contextual factors may have worked to erode the traditional Republican advantage on taxes. In 1992 Clinton successfully challenged Bush on the basis of a poorly performing economy, effectively positioned himself as a "centrist" following the Republican gains in 1994, and then presided over a rapidly recovering economy. Based on his recent economic accomplishments, we might predict that in 1996 Clinton could temporarily "lease" the issue of taxes from the Republicans (see Petrocik, 1996).

In summary, we expect that Democratic candidates will benefit when the subject of the ad campaign is unemployment, womens' rights, education, or government funding of social welfare programs. Conversely, we anticipate that Republicans will enjoy greater credibility when the campaign issue agenda consists of crime or illegal immigration. Given the recent history of the issue, we make no prediction on the issue of taxes.


Our first test of credibility as a mediator of campaign advertising analyzes the electoral support for the candidates in relation to the issue agenda established by the advertising campaign. As shown in Figure 1, the candidates' electoral prospects were enhanced when the issue agenda of campaign advertisements matched their party affiliation.

(Figure 1 About Here)

The first panel in Figure 1 displays the results for the 1992 California Senate campaigns. In these elections, support for the Democratic candidates increased by nearly 7 percent when the agenda switched from Republican (crime) to Democratic (employment or gender-related) issues. When respondents were exposed to ads about crime, support for the Democratic candidates was 63.2 percent, no different from the level among those who saw no political ad at all. This finding suggests that the Republican candidates in these particular races were unlikely to win under any circumstances, but their prospects significantly worsened when respondents saw ads involving women's rights or unemployment.

A similar pattern emerged in the 1994 gubernatorial campaign. As displayed in Panel B of Figure 1, when both candidates addressed education and jobs, Democrat Kathleen Brown won handily (with 58.5 percent of the two-party vote). When the issue agenda was mixed and each candidate advertised on his or her strengths, the race was a standoff (Brown received 53 percent of the vote). Finally, when both campaigns advertised on crime or immigration, Wilson was the decisive victor (receiving 56 percent of the vote). Clearly, the candidates' choice of issues was crucial to the outcome of the race. Had economic issues predominated, Brown might have been elected; however, the Wilson campaign was able to tap the public's fear of crime and illegal immigration and thereby succeeded in pushing economic issues to the background.

The results from the 1996 presidential race further reinforce the notion that a candidate benefits when the central issues of the ad campaign are `owned' by his or her party. As illustrated in the third panel of Figure 1, among participants exposed to ads about the future of Social Security, Medicare, college loans and job training, Clinton enjoyed widespread support (62.6 percent). On the other hand, Clinton's vote share was weakest (at 56.1 percent) when respondents were exposed to advertisements about the traditional Republican issues of illegal immigration or drug-related crime. Surprisingly, participants exposed to ads about taxation were just as likely to vote for Clinton as those who watched ads dealing with social welfare. These results suggest either that "cutting taxes" is no longer an inherently Republican appeal or that there is a centrist trend in public opinion that resists large shifts in the economic status quo (as represented by the Dole tax cut proposal). The relatively strong performance of the economy coupled with the radical nature of the Dole tax plan may have combined to grant Clinton special credibility on the tax issue in 1996.

To further test these differences, we controlled for party identification, gender, education, and respondents' previous presidential vote. Tables 1 and 2 present these multivariate tests.

(Tables 1 and 2 About Here)

The dependent variable in these analyses, as mentioned before, is the percentage point lead for the Democrat. A trichotomous variable was created to measure the issue agenda in each campaign. Messages focusing on Republican issues were coded -1, while those focusing on Democratic issues were coded +1. Exposure to messages involving taxation, or exposure to no political advertisements at all, was coded equal to 0. Thus a positive coefficient on the issue agenda variable corresponds to the average increase in the lead for the Democratic candidate associated with moving from Republican to neutral issues, or from neutral to Democratic issues. In the first row of Table 1, we see that the Democratic lead increased by an average of 6% for each level of the trichotomous variable corresponding to the issue agenda of the advertising campaign, even after controlling for various alternative influences on vote choice. Thus moving from Republican issues to Democratic issues led to a 12% increase in support for the Democrat.

The effect of the issue agenda was quite similar in 1996 and was also statistically significant. The coefficient of .05 indicates that moving from Republican issues (immigration or drugs) to Democratic issues (spending on social programs) boosted the Democrat's support by 10%.

Table 2 presents the results from the 1994 California gubernatorial race. Even with the introduction of several control variables, Kathleen Brown's lead surged significantly when campaign advertising focused on Democratic issues. Each level of the campaign agenda variable was associated with an increase of 11 percentage points in support for Brown. Thus, when both candidates focused on education or unemployment rather than crime or illegal immigration, the Democrat improved her electoral support by an impressive 22%. Overall, these results confirm the pattern revealed in Figure 1; shifting the issue agenda of the campaign significantly affects the fortunes of the candidates.

To this point we have been concerned with the issue agenda of campaign advertising, without regard to the specific source of advertisements. But in order to test precisely the resonance model of campaigns, we must look separately at voter reactions to issue appeals by candidates from each party. Our second test, therefore, considers the credibility of campaign advertising as a function of both message content and source. We examined differences in voter choice when particular candidates aired advertisements dealing with particular issues. Does a Democratic advertisement on unemployment sway more voters than a Democratic advertisement on crime? Conversely, are Republicans more effective with advertisements on crime than unemployment?

This test requires a one-ad design in order to isolate the effects of individual candidates and individual issue appeals. Therefore, we can present evidence only from the 1992 and 1996 campaigns. Figure 2 displays the mean level of support for Democratic candidates for each level of the issue agenda established by each of the sponsoring candidates.

(Figure 2 About Here)

In 1992, both Democratic senatorial candidates improved their prospects by 9 percentage points when they advertised on job creation and women's rights, and actually worsened their chances (by two points) when they advertised on crime. The Republicans were generally ineffective no matter what issue they used, as indicated by the almost identical levels of support for the Democrat across the advertising categories in Panel A of Figure 2.

Turning to the 1996 campaign, Clinton maximized his vote share, (64.9 percent) when he advertised on social welfare programs and minimized his vote share (55%) when he addressed the Republican issues of illegal immigration and drugs. Conversely, Dole's most successful messages involved immigration and drugs. The issue of taxes appeared to favor Clinton in that he maintained his commanding lead over Dole when advertising on this issue, while Dole's tax-related ads worsened his position considerably. Overall, the evidence in Panel B of Figure 2 indicates that Clinton was invulnerable so long as he did not stray into Republican territory. For Dole, attacks on Clinton as a tax-and-spend liberal and appeals on the basis of his tax plan only played into Clinton's hands. Dole could only be competitive when his advertisements centered on drugs or immigration.

We again tested the statistical significance of these differences in a regression analysis that included multiple controls. These results are presented in Table 3. In 1992, Republicans were unable to improve their vote share significantly regardless of their choice of advertising messages. The Democrats, however, could boost their lead substantially by advertising on "their" issues. The 11 point shift associated with each level of the trichotomous campaign agenda variable translates into a 22% increase in support for the Democrats when they chose to advertise on employment or women's issues instead of crime.

In 1996, the Democrat's choice of ads again significantly influenced his lead. Shifting from immigration and drugs to spending on social programs produced a 12 percentage point (.06 x 2) increase in Clinton's support. Conversely, Dole's optimal result was when he advertised on crime and immigration, though the effect was not significant. When Dole's ads concerned social welfare issues, his level of support fell by 6 percent. In effect, any Dole ad not dealing with drug abuse or immigration only served to worsen his position vis-a-vis Clinton.

(Table 3 About Here)

In summary, we have demonstrated that issue content is a significant determinant of the persuasiveness of campaign advertising. The credibility of political advertising is enhanced when candidates resort to traditional issue appeals -- appeals on which their party has a longstanding advantage in the minds of voters. We can test this hypothesis even more precisely in 1996 because in that study we included items that directly tap viewers' beliefs about the credibility of televised political advertising. In particular, we asked respondents to indicate how well the terms "informative" and "misleading" applied to "commercials for political candidates or organizations." Responses to these items were used to construct an index of credibility which runs from -1 (uninformative and misleading) to +1 (informative and not misleading). Because we insisted that respondents evaluate the believability of political advertising in general and not the specific ads they had just seen, this is an especially strict test of the credibility hypothesis. Table 4 presents these results.

Panels A and B of Table 4 presents the mean credibility scores for each sponsoring candidate and the issue content of their messages. There are three levels of issue content corresponding to "owned issues" (issues on which the sponsoring candidate's party is favored), "opponent's issues" (issues on which the opposing candidate's party has the edge) and "ambiguous issues" (on which neither party may have a clear advantage). For Dole, ads dealing with crime and immigration fall into the owned issue category, ads on taxes are classified as ambiguous with respect to ownership, and advertising on the budget is considered trespassing into Democratic territory. Of course, the reverse logic would apply to advertising by Clinton.

Before addressing the effects of issue content and partisanship of the sponsoring candidate on the credibility index, it is noteworthy that campaign advertising was rated much more negatively than product advertising. The sample mean for campaign advertising was -.39 indicating that a large plurality of participants rated political ads as lacking in credibility. For product advertising, the sample mean was -.03, indicating that roughly equal numbers of participants rated product ads as informative and truthful as uninformative and misleading. From the perspective of the audience, campaign advertising is clearly tainted.

The results in Figure 4 accord well with our previous findings. In Panel A and B, credibility was highest when Dole advertised on traditional Republican issues and when Clinton's appeal was consistent with his party affiliation. Although none of the observed differences were statistically significant, they at least suggest that political advertising is more likely to be considered credible when the message and messenger match.

(Table 4 About Here)

Since the theoretical notion of credibility can be generalized across individual candidates (and because the pattern of differences noted above was so similar for both Dole and Clinton), we pooled across candidates and subjected the credibility index to a multivariate analysis that controlled for individual differences in attitudes toward campaign advertising. Panel C of Table 4 presents the adjusted credibility means within each level of issue content. As anticipated, the effects of issue content were significant -- moving from the "owned issue" to "opposition party's issue" category yielded a .10 downward shift (p < .05) in the average level of credibility. This finding strongly supports our original hypothesis: respondents rate advertisements as significantly less credible when the sponsor's party and the issue focus of the ad are incongruent.

In summary, our evidence reveals that campaign advertising is most effective when the sponsoring candidates pitch their message to the traditional strengths of their parties. Voters tend to place their credence in advertisements that correspond to the sponsor's partisan reputation. This pattern holds for senatorial, gubernatorial, and presidential campaigns. Ceteris paribus, the candidate whose issue agenda fits better with voters' stereotypes about the capabilities of the political parties will carry the day.


Our evidence supports the resonance model of campaigns. A candidate's party conveys information about his or her ability to deal with particular issues. Our results suggest quite strongly that candidates should seek to exploit such reputational advantages. A Democrat who calls for educational reform or for more stringent enforcement of gender discrimination laws will be taken more seriously than a Democrat who favors the death penalty or more aggressive monitoring of terrorist groups.

An important implication of these findings is that in competitive races, where both candidates tend to be well-financed, the total amount of advertising purchased by a candidate may be secondary to the selection of the appropriate message. In both the Senate and presidential studies, the sponsoring candidate who chose to advertise on the "wrong" issue was better off showing no advertisement at all. This pattern suggests that it is the issue agenda of the campaign rather than the volume of each candidate's campaign communication that influences voters.

Another important implication is that campaigns are unlikely to provide meaningful dialogue about the important issues of the day. Our results suggest that candidates have a strong incentive to focus on issues that resonate with their party affiliation, no matter what the opposition chooses to discuss. Though candidate dialogue (both candidates addressing the same issues) is undoubtedly the best way for voters to make enlightened choices, candidates cannot be expected to sacrifice their own electoral chances by engaging the opposition on issues they place them at a disadvantage. In the long run, this strategy contributes to the confusion and tumult of campaign communication, and to the dissatisfaction voters feel with the candidates (for a general model of campaigns that predicts the absence of dialogue, see Simon, 1997).

The results provided by our experiments are probably conservative estimates of the importance of source credibility in campaign advertising since all the elections under consideration were relatively "high stimulus" races characterized by well-known and well-financed candidates whose campaigns generated considerable news coverage and publicity in addition to paid advertising. When voters encounter advertising from less visible candidates, the importance of the candidates' party affiliation is likely to be amplified. In general, we might expect that source-related cues increase in importance as voters' store of information about particular candidates declines.

The use of the sponsoring candidate's attributes as a basis for inferring his or her positions on political issues extends well beyond the candidate's party affiliation. Gender and race are especially visible attributes and the popular culture provides several associations between these traits and political ideology (see McDermott, 1997). Thus, in the case of gender, messages involving stereotypically "masculine" issues such as crime or national defense should be especially persuasive when the sponsor is a male war hero. On the other hand, child care and nutrition will fit well with voters' beliefs about the credentials of a candidate who happens to be a mother. Ansolabehere and Iyengar (1994) and Iyengar et al. (1996) have assessed the relative credibility of male and female candidates on the issue of sexual harassment. Their results showed that advertising on this issue yielded significant gains for both women running for the U.S. Senate in 1992, but virtually no return for presidential candidate Clinton.

In conclusion, our evidence in support of the resonance model suggests once again that in low-involvement persuasion situations it is the interplay between "old" and "new" information that governs the effects of mass communication. Political campaigns are classic low-involvement situations: most viewers are less than fully captivated by political campaigns and pay only casual attention to campaign advertisements. Given the fleeting nature of most voters' exposure to campaign advertising, it is the advertiser's ability to evoke what voters already "know" about the candidates that makes the difference between effective and ineffective advertising campaigns.


Ansolabehere, S., and S. Iyengar. 1994. Riding the Wave and Exercising Ownership Over Issues. Public Opinion Quarterly, 58, 335-57.

Baumer, D. L., and H. J. Gold. 1995. Party Images and the American Electorate. American Politics Quarterly, 23, 33-61.

Buchanan, Bruce. 1991. Electing a President: The Markle Commission's Report on Campaign '88. Austin: University of Texas Press.

Budge, I., and D. Fairlie. 1983. Explaining and Predicting Elections. London: Allen and Unwin.

Goldstein, D. 1997. Political Advertising and Political Persuasion in the 1996 Presidential Campaign. Paper Presented at the Annual Meeting of the American Political Science Association.

Hovland, C. 1959. Reconciling Conflicting Results from Experimental and Survey Studies of Attitude Change. American Psychologist, 14, 8-17.

Iyengar, S. 1996. The case of the vanishing footprints: A review of research on political campaigns. Paper Presented at the Annual Meeting of the Association for Education in Journalism.

Iyengar, S., N. Valentino, S. Ansolabehere, and A. Simon. 1996. Running as a woman: Gender stereotyping in women's campaigns, in P. Norris ed., Women, Media, and Politics. New York: Oxford University Press.

Jennings, M. K., and R. G. Niemi. 1981. Generations and Politics: A Panel Study of Young Adults and Their Parents. Princeton: Princeton University Press.

Kleinnijenhuis, J., and J. de Ridder. 1997. Effects of issue priorities in the news on voting preferences, in C. W. Aarts and B. Steenge eds., Setting Priorities in the Public Sector. Dordrecht, Netherlands: Kluwer.

Lupia, A. 1997. Who Can Persuade Whom? How Simple Cues Affect Political Attitudes, in A. Lupia and M. McCubbins eds.,

McDermott, M. 1997. Voting Cues in Low-Information Elections: Candidate Gender as a Social Information Variable in Contemporary U.S. Elections. American Journal of Political Science, forthcoming.

McGuire, W. J. 1985. Attitudes and attitude change, in G. Lindzey and E. Aronson eds., Handbook of Social Psychology. New York: Random House.

Petrocik, J. 1996. Issue Ownership in Presidential Elections with a 1980 Case Study. American Journal of Political Science, 40, 825-50.

Petty, R., and J. Cacioppo. 1984. Attitudes and Persuasion: Classic and Contemporary Approaches. Dubuque, Iowa: W. C. Brown.

Popkin, S. L. 1991. The Reasoning Voter: Communication and Persuasion in Presidential Campaigns. Chicago: University of Chicago Press.

Sears, D.O. 1983. The persistence of early political predispositions: the roles of attitude object and life stage, in L. Wheeler & W.J. McGuire eds. Political Socialization, Citizenship Education, and Democracy. New York: Teachers College Press.

Shaw, D. 1997. Candidate Appearances and Television Advertising in the 1988-1996 Presidential Elections. Paper Presented at the Annual Meeting of the American Political Science Association.

Simon, A. 1997. The Winning Message. Ph.D. Dissertation, University of California, Los Angeles.
 Table 1

The Effects of the Campaign Issue

Agenda on Vote Choice





California Senate Races 


Presidential Race

Issue Agenda*



Party Identification









Previous Presidential Vote






N 953 1034


Table 2

The Effects of Owned Versus Mixed Campaign

Issue Agendas on Vote Choice in 1994

1994 Gubernatorial Race

Campaign Issue Agenda


Party Identification




Vote on Prop. 187


Previous Presidential Vote








  Table 3

The Effects of the Campaign Issue Agenda

on Vote Choice By Sponsoring Candidate














Issue Agenda -.02 .11*** .03 .06*
Party Identification















Previous Presidential Vote















N 447 677 733 485

*Note: The dependent variable is the Democratic candidate's percentage point lead in the major party vote. Entries are unstandardized OLS regression coefficients. Issue agenda is coded ?1 for exposure to ads on Republican issues (crime in 1992 and immigration/drugs in 1996), 0 for exposure to ads dealing with issues on which neither party could claim outright ownership (taxes in 1996), and +1 for exposure to ads on Democratic issues (unemployment and women's rights in 1992, social welfare programs in 1996). Party identification is coded Democrat = 1, Republican = -1, Independent = 0. Gender is a dummy with Female = 1. Education is coded from 1 = some high school through 5 = post-college. Previous presidential vote is coded Bush = -1, Clinton = 1, Other=0. * p # .10, ** p # .05, *** p # .01.


 Participant Profile






Median Age




Percent With College Degree 53 54 50
Party Identification      
24 31 33
31 32 26
45 38 39
7 15 4
20 22 8
8 12 8
62 51 75