Political knowledge and sponsorship in backlash from party- and candidate-sponsored attacks.

Apparently wary of backlash, candidates are sponsoring fewer negative ads while their party's national committees do more of the dirty work. In this experiment (N = I07), contrary to expectations, candidate evaluations were more positive when participants saw the candidates' own attacks than

when they saw party-sponsored attacks. An interaction showed that this was true particularly for viewers higher in political knowledge in their overall candidate evaluations. The recent embrace of 'plain-spoken' candidates is mentioned as a possible explanation. Implications of the Bipartisan Campaign Reform Act in light of this study are discussed.

Keywords: Negative Political Advertising; Political Knowledge; Sponsorship; Political Campaigns

**********

In political advertising, backlash is a well-documented phenomenon in which a political attack ad ends up lowering evaluations of the candidate who sponsored it, often as much or more than the candidate criticized in the ad (Garramone, 1984; Haddock & Zanna, 1997; Hill, 1989; Jasperson & Fan, 2002; Lau, Sigelman, Heldman, & Babbitt, 1999; Merritt, 1984; Shen & Wu, 2002).

Apparently wary of backlash, candidates in recent years have run fewer negative ads themselves and left more of the dirty work to their party's national committees or to political action committees (Stanger & Rivlin, 1998). In the 2000 US federal elections, nearly half of party ads were exclusively negative, compared to just 16% for candidate-funded ads (Holman & McLoughlin, 2002). 'Apparently, without a specific name of a person behind the ad, parties and groups feel freer to go negative and attack candidates on their merits or character' (Holman & McLoughlin, 2002, p. 66). Candidates may hope that the attacks by proxy waged by their political allies will not lead to a backlash against themselves.

Despite the fact that independent ad expenditures have overtaken candidate ad expeditures (Marcus, 2000), party-sponsored political advertising has received 'scant attention in political communication research' (Pfau, Holbert, Szabo, & Kaminski, 2002, p. 302). The topic has become particularly relevant with the ongoing legal wrangling over campaign finance reform. The Bipartisan Campaign Reform Act (2002) set out to ban the raising and spending of 'soft money,' unregulated contributions by individuals, unions and corporations to the national party committees. A US District Court in May 2003 struck down the soft money ban but left intact a provision requiring parties to pay hard money for any television ad that refers to a specific candidate.

One rationale offered for limiting such ads is their presumed influence on voters. But that presumed influence has received little study. Likewise, there is little research on whether candidates incur less backlash from party-funded negative ads than their own. Garramone (1985) found evidence that televised attacks from independent sponsors are more successful than attacks from a candidate. However, Garramone's work came well before the explosion of issue advertising by parties and political action committees (PACs) since 1996, and her experimental stimuli made sponsor information much more prominent than it has been in recent years.

Some recent work has examined this question. Using real-world television ads for real candidates as experimental stimuli, Pfau and colleagues (2002) found no main effects for sponsor or tone. Shen and Wu (2002) found only moderate evidence of backlash upon a candidate whose opponent is attacked by an independent ad. Overall candidate perceptions and character perceptions were significantly lower for those viewing such ads than for those in a control group, but candidate liking and voting intentions were not. Shen and Wu had hypothesized no backlash because 'it is usually hard for a layperson to detect the hidden links between the political organizations and the candidates they implicitly support' (p. 399). They performed no direct comparisons between candidate- and independent-sponsored negative ads, but they found less evidence of backlash in the case of independently-sponsored ads.

However, Jasperson and Fan's (2002) case study of advertising tone, frequency, and favorability polls in the 1996 US Senate race in Minnesota found that the Republican challenger's own attack ads did not appear to lower his own evaluations, but party-sponsored attacks on his behalf did. They attributed this counterintuitive finding to the fact that the Republican party-sponsored attacks contained more negative information than the candidate's ads, many of which were comparative rather than purely negative. Another consideration may be that the challenger, Rudy Boschwitz, was well-known and had held the Senate seat for 12 years before losing it in 1990. Under such circumstances, perhaps it was not difficult for voters to connect Boschwitz to the party sponsoring the attacks. The rationale Shen and Wu (2002) offered for their prediction of no backlash for such ads, that it is harder to connect a candidate with an attack in such cases, makes sense here, and may explain why Jasperson and Fan found the opposite result in their study.

H1: Evaluations of a candidate will be more negative when the candidate sponsors an attack than when the candidate's party does.

If backlash requires the voter to connect a candidate with an attack, then political knowledge would play a key role. Without some knowledge, people are unable to make political decisions that reflect their interests and values (Delli Carpini & Keeter, 1996). Political knowledge helps people translate their interests and values into appropriate policy preferences, and in turn recognize candidates whose policies better match their own (Meirick & Wackman, 2004). In the context of political advertising, we would expect that those with higher political knowledge, because of their ability to connect an attack to the candidate on whose behalf it is made, to have lower evaluations of that candidate.

H2: Political knowledge will be negatively related with evaluations of the candidate on whose behalf an attack runs.

Political knowledge should play a role in sponsor recognition, although this has received little study. Those high in political knowledge are likely to recognize photos and videos of the candidates, know what offices they are seeking and whom they are running against. Similarly, they are likely to be aware of the prevalence of party- and PAC-sponsored advertising and be on the lookout for it.

H3: Political knowledge will be positively related to correct sponsor identification.

The final hypothesis deals with political knowledge as a moderating variable in the workings of backlash. Hypothesis 2 specifies that political knowledge allows people to connect campaign information with the correct candidates, so we expect those with higher knowledge to have lower evaluations of candidates whose campaigns or parties make attacks. But, as Hypothesis 3 argues, political knowledge also should facilitate the ability to differentiate between candidate-sponsored and party-sponsored ads. If so, the differences in candidate evaluations between these types of ads that are expected in Hypothesis 1 should be more pronounced among those higher in political knowledge.

H4: Political knowledge and sponsor will interact such that knowledge will be more negatively related with candidate evaluations for participants who see a candidate-sponsored ad rather than a party-sponsored ad.

Method

Participants were recruited from communication classes at a university in the Southwest. In all, 55 men and 52 women participated. The mean age was 20.2. About 20% identified themselves as members of racial or ethnic minorities. Initially, to avoid sensitizing participants to the political advertising they would see, the participants were told only that the study concerned television viewing and attitudes. Participants took part in the study one to five days before the 2002 general election. As they arrived, they were greeted and randomly assigned to one of four conditions, representing a 2 (candidate- or party-sponsored) x 2 (candidate who benefited from the ad: Republican or Democrat) design: Democratic candidate-funded attack ad (n = 26), Democratic party-funded attack ad (n = 27), Republican candidate-funded attack ad (n = 26), or Republican party-funded attack ad (n = 28). The ads were imbedded in a 15-minute segment of a game-show program during which the other advertising in two ad breaks was identical.

The ads used in this study were part of a campaign for an open Congress seat in the district where the university is located. They were taped off the air during the month leading up to the general election. Because these were actual ads, experimental control could not be exercised over all message elements. However, message coordination between the parties and candidates was evident, and ads were chosen so that a candidate's ads had the same themes. The party- and candidate-sponsored ads on behalf of the Republican candidate both attacked the Democratic candidate for a record of raising taxes. The party- and candidate-sponsored ads on behalf of the Democratic candidate both attacked the Republican candidate for not supporting education funding and for not serving in the military.

Participants were each given a videotape containing one of the four versions of the stimuli and taken to a room containing seven 13-inch TV/VCRs in private study carrels, each with a set of headphones. After watching the stimuli, participants returned and were given the appropriate questionnaire. The study took most participants 30 minutes to complete.

Measures

Political knowledge

A 13-item scale including four open-ended questions about officeholders (e.g., who is the vice-president?), three open-ended questions about the system (e.g., how many justices are on the Supreme Court?), and six dichotomous-choice questions about party positions (e.g., which party tends to oppose abortion more?). The scale had acceptable reliability (Cronbach's [alpha] = .74). Mean score on the knowledge scale was 7.4 (SD = 2.95). Scores ranged from 1 to 13.

Subject's partisanship

This item used a seven-point scale from 1 (strong Republican) to 7 (strong Democrat), with a midpoint of 4 (independent, not closer to either party). The mean score was 3.4 (SD = 1.61).

Sponsor identification

The questionnaire told participants that they had a 'seen an ad that criticized a candidate for the US House of Representatives in District 4.' Participants were asked, 'What group or campaign do you think paid for that ad?' Answer options were 'no idea,' 'another candidate,' 'a political party,' and 'a PAC or some other group.' They were then asked to specify what party, candidate, or PAC, if they could, in an empty blank. These two items were scored as correct or incorrect and combined to create a Guttman-type scale with three levels (both wrong = 0; the less specific item right = 1; both items right = 2). The coefficient of reproducibility was 1.00; for instance, no one correctly named a sponsoring candidate after incorrectly identifying a political party as the sponsor of the ad.

Candidate traits

Perceptions of candidates' competence and character were assessed using bipolar adjective scales developed by McCroskey and Jenson (1973) and employed in the work of Pfau and colleagues (Pfau & Burgoon, 1988; Pfau et al., 2002). The adjective pairs for competence were unintelligent-intelligent, unqualified-qualified, and incompetent-competent. For character, the pairs were dishonest-honest, selfish-unselfish, and bad-good. Each was item scored on a 9-point scale. Alpha reliabilities for competence were .91 for the Republican candidate and .83 for the Democratic candidate; for character, alphas were .89 and .86, respectively.

Overall evaluation

Participants' overall attitude toward the candidates was measured with a scale adapted from Burgoon, Cohen, Miller, and Montgomery (1978). As above, the scale consisted of three adjective pairs with a 9-point scale for each item: unacceptable-acceptable, unfavorable-favorable, and wrong-right. Alpha reliabilities were .93 for the Republican candidate and .90 for the Democrat.

Results

Hypothesis 1 predicted that evaluations for candidates would be lower when they sponsored attack ads than when their party did. A MANOVA was run with competence, character, and overall evaluation as dependent variables. There were no significant interactions between candidate and ad sponsor, the independent variables. Candidate made a difference, with the eventual winner (the Republican) rating more highly. More to the point, sponsorship mattered as well, but in an unexpected way--candidates were rated more highly across the board when participants saw them sponsoring their own attacks rather than party-sponsored attacks, multivariate F(3, 92) = 2.789, p < .05 (see Table 1). Hypothesis 1 is contradicted.

Hypothesis 2 predicted that political knowledge would be negatively related to evaluations of the candidate on whose behalf an attack runs. To test this hypothesis, zero-order correlations were run. To control family-wise error rate, the alpha level was set at .0167 (.05/3). Contrary to the hypothesis, knowledge was positively but nonsignificantly correlated with evaluations of character, r(98) = .21, n.s., competence, r(100) =. 11, n.s., and overall evaluation, r(101) = .13, n.s. Hypothesis 2 is not supported.

The third hypothesis predicted that political knowledge would be positively related to correct sponsor identification, and it was, r(105) = .27, p < .001, one-tailed. However, this result does not account for other variables. In a multiple regression that added previous exposure to the ad, sponsor type (candidate- or party-sponsored), and candidate (Republican or Democrat), political knowledge remained a significant predictor, [beta] = .23, t(106) = 2.46, p < .05. Hypothesis 3 is supported.

Hypothesis 4 predicted that political knowledge and sponsor would interact such that those with higher knowledge would be more likely than others to negatively evaluate candidates who sponsored attack ads themselves rather than rely on a party-sponsored ad. Three multiple regression equations were used, with candidate evaluations (character, competence, and overall) as the dependent variables. Exogenous variables included candidate (Republican or Democrat), ad sponsor (candidate or party), subject's party lD, gender, race (Caucasian or not), and political knowledge. These variables were centered to avoid collinearity problems with the interaction term (Yu, 2000), which was the key test of the hypothesis: the product of political knowledge and ad sponsorship. Using multiple regression avoided constraining political knowledge and party ID to two or three levels and allowed a closer examination of the findings under Hypothesis 1.

Table 2 shows that, as before, the Republican candidate was evaluated more highly, while Democratic respondents had lower overall evaluations of the attacking candidate. The role of ad sponsorship remained robust; evaluations were more negative with party-sponsored attacks, again contrary to Hypothesis 1.

As for the current hypothesis, results were mixed and, again, tended toward contradiction. Given the valence of the variables, the predicted interaction would be reflected in a significant positive coefficient for the interaction term such that evaluations would be greater when knowledge was higher and ads were part? sponsored. However, the coefficients of the interaction terms were negative in each case, significantly so for overall evaluation. The nature of the interaction is shown in Figure 1. Those low in knowledge evaluated the candidate essentially the same regardless of who sponsored the attack, while those high in knowledge evaluated the candidate more highly when the candidate sponsored his own attack. This hypothesis is not supported, and some evidence suggests contradiction.

[FIGURE 1 OMITTED]

Discussion

Attacking candidates were evaluated more favorably when participants saw candidate-sponsored attacks rather than part?-funded attacks, contrary to expectations and to the common practice of letting parties and PACs run most attacks. As expected, political knowledge was positively associated with sponsor recognition of candidate- and party-funded ads, and it played a role in overall evaluations of attacking candidates, but in an unexpected way. Knowledge interacted with ad sponsorship such that among those with higher knowledge, overall evaluations were higher when the ad was candidate-sponsored and lower when part?-sponsored.

Why did candidate-sponsored ads lead to more positive evaluations? In recent years, the public and the media have responded favorably (at least initially) to 'plain-talking' politicians such as Jesse Ventura, John McCain, and Howard Dean, as well as the fictional Pres. Josiah Bartlett on NBC's 'The West Wing.' These politicians are prone to blunt assessments of their opponents. To voters, such tendencies may be indicative of honesty, forthrightness, and directness. This explanation is speculative, but in this study, evaluations of a candidate's honesty were significantly higher for candidate-sponsored (M = 5.38) than for party-sponsored ads (M = 4.68), t(101) = 2.185, p < .05.

One alternative explanation that some might offer is that candidate-sponsored attacks may be less negative than party-sponsored attacks because candidates want to avoid backlash, and greater levels of negativity have been associated with lower candidate evaluations (Pinkleton, 1997). It is true that the Republican candidate ran a comparative ad while the RNC's ad was purely negative. But the Democratic candidate's own ad was much harsher than his party's ad. In his ad, the Democratic candidate himself said of his opponent, 'He dodged the draft, enrolling in graduate school to get deferment after deferment;' his party's ad said only, 'He didn't serve.' Moreover, there were no candidate-sponsor interactions; the pattern of preference for candidate-sponsored ads held for both candidates.

If the public is souring on party-sponsored attacks and is willing to see candidate-sponsored attacks as evidence of virtue, it appears to be especially true of those high in political knowledge. The logic that led to the final hypothesis--that the high-knowledge contingent would be better able to connect information and inferences from an ad to a specific candidate--serves just as well to explain why the hypothesis was partially contradicted. The only difference is that the inferences they made from candidate-sponsored attacks were more positive than expected, and inferences from party-sponsored attacks were more negative.

Another part of the rationale for the final hypothesis was that sponsor information was purposely obscured in most attack ads. However, sponsor information in political advertising has become more prominent as a result of the BCRA's 'Stand By Your Ad' provisions. To comply with them, candidates must appear on camera identifying themselves and saying, 'I approved (or authorized) this message.' Third-party advertising must include a similar declaration from a representative of the sponsoring group (Koerner, 2003). If the old subtle sponsor cues meant that the only politically knowledgeable were likely to appreciate a candidate's forthrightness in doing his own attacking, perhaps the new clarity standards and the obvious sponsor cues will broaden this effect to all levels of political knowledge. If so, candidates might do well to stop delegating attacks to parties and PACs and do it themselves.

This study has several limitations. The sample was small and was drawn by convenience from a student population. This limitation is especially important to note for Hypotheses 2 and 3, which do not benefit from the experimental logic of random assignment. The number of ads represented (four) was small, and the number of candidates (two) and races (one) smaller still. The ads and candidates were real, so they were not subject to precise control. The absence of a control group means that inferences about the absolute (rather than relative) effects of attack sponsorship cannot be made. Overall, we should be cautious in making generalizations from this study about the effects of sponsorship on reactions to attack ads; these findings are not conclusive. Still, the limitations are not grave. First, while we should not take the means found here as US population parameters, there is no clear reason to believe that the relationships between knowledge, sponsorship, and candidate evaluations would not occur in a more representative population. Second, experiments with small numbers of ads are common in political advertising research, and the results here were consistent across candidates. Finally, the fact that the study used real ads should bolster external validity.

Future research should follow up on the potential effects of obvious sponsor cues and see if the knowledge-sponsorship interaction is attenuated. Measures that specifically and reliably tap perceptions of 'plain-spokenness' as a dimension of candidate evaluation should be developed to see how ad sponsorship affects it. Finally, sponsorship and tone of advertising could be tracked and consultants interviewed to see if the conventional wisdom about the consequences of negativity changes, and if candidates get their hands dirtier.

References

Bipartisan Campaign Reform Act of 2002, Pub. L. No. 107-155, [section] 2, 116 Stat. 81-116. Retrieved Oct. 17, 2003, from: http://www.fec.gov/pages/bcra/major_resources_bcra.htm

Burgoon, M., Cohen, M., Miller, M. D., & Montgomery, C. L. (1978). An empirical test of resistance to persuasion. Human Communication Research, 5, 27-39.

Delli Carpini, M. X., & Keeter, S. (1996). What Americans know about politics and why it matters. New Haven, CT: Yale University Press.

Garramone, G. M. (1984). Voter response to negative political ads. Journalism Quarterly, 61 (2), 250-259.

Garramone, G. M. (1985). Effects of negative political advertising: The roles of sponsor and rebuttal. Journal of Broadcasting and Electronic Media, 29 (2), 147-159.

Haddock, G., & Zanna, M. P. (1997). Impact of negative advertising on evaluations of political candidates: The 1993 Canadian federal election. Basic and Applied Social Psychology, 19 (June), 204-223.

Hill, R. P. (1989). An exploration of voter responses to political advertisements. Journal of Advertising, 18 (4), 14-22.

Holman, C. B., & McLoughlin, L. P. (2002). Buying time 2000: Television advertising in the 2000 federal elections. New York: Brennan Center for Justice at NYU School of Law.

Jasperson, A. E., & Fan, D. P. (2002). An aggregate examination of the backlash effect in political advertising: The case of the 1996 U.S. Senate race in Minnesota. Journal of Advertising, 31 (1), 1-12.

Koerner, B. I. (2003, July 28). Why all the stilted new campaign ads? How McCain-Feingold will shape candidates' spots. Slate. Retrieved Oct. 29, 2003, from http://slate.msn.com/id/2086194/

Lau, R. R., Sigelman, L., Heldman, C., & Babbitt, P. (1999). The effects of negative political advertisements: A meta-analytic assessment. American Political Science Review, 93 (4), 851-875.

Marcus, R. (2000, November 6). Costliest race nears end; Bush, Gore running close; U.S. campaigns fuel $3 billion in spending. The Washington Post, p. A1.

McCroskey, J. C., & Jenson, T. A. (1973). Image of mass media news sources. Journal of Broadcasting, 19, 169-180.

Meirick, P. C., & Wackman, D. B. (2004). Kids voting and political knowledge: Narrowing gaps, informing votes. Social Science Quarterly, 85 (5), 1161-1177.

Merritt, S. (1984). Negative political advertising: Some empirical findings. Journal of Advertising, 13 (Fall), 27-38.

Pfau, M., & Burgoon, M. (1988). Inoculation in political campaign communication. Human Communication Research, 15, 91-111.

Pfau, M., Holbert, R. L., Szabo, E. A., & Kaminski, K. (2002). Issue-advocacy versus candidate advertising: Effects on candidate preferences and democratic processes. Journal of Communication, 52 (2), 301-315.

Pinkleton, B. (1997). The effects of negative comparative political advertising on candidate evaluations and advertising evaluations: An exploration. Journal of Advertising, 26 (Spring), 19-29.

Shen, F., & Wu, H. D. (2002). Effects of soft-money issue advertisements on candidate evaluation and voting preference: An exploration. Mass Communication and Society, 5 (4) 395-410.

Stanger, J. D., & Rivlin, D. G. (1998). Issue advocacy advertising during the 1997-1998 election cycle. Philadelphia: University of Philadelphia, Annenberg Public Policy Center.

Yu, C. H. (2000). An overview of remedial tools for collinearity in SAS. In Proceedings of 2000 Western Users of SAS Software Conference (pp. 196-201). Scottsdale, AZ: WUSS.

Patrick C. Meirick (PHD, University of Minnesota, 2002) is an Assistant Professor of Communication at the University of Oklahoma. Correspondence to: Patrick C. Meirick, Department of Communication, University of Oklahoma, 610 Elm Ave., Norman, OK 73019, USA. Email: meirick@ou.edu

Table 1 Attack Ad Sponsorship and Evaluations of the
Attacking Candidate

                        Sponsor of attack ad

Type of evaluation   Party   Candidate    F value

Competence           15.79     18.02      6.315 *
Character            14.46     16.58      6.721 *
Overall              14.58     16.82     7.847 **

Note. Evaluation scores are sums of three 1-9 polar adjective
items. N = 100-103.

* p < .05. ** p < .01. *** p < .001.

Table 2 Predictors of Candidate Evaluations

Predictor                          Competence   Character   Overall
                                    ([beta])    ([beta])    ([beta])

Race (Caucasian) (a)                 .12          .21 *       .08
Gender (female)                     -.08          .05        -.06
Candidate (Democrat)                -.33 ***     -.28 **     -.34 ***
Ad sponsor (party)                  -.23 *       -.23 *      -.24 **
Respondent's party ID (Democrat)    -.21         -.08        -.21 *
Political knowledge                  .04          .16         .06
Political knowledge x ad sponsor    -.13         -.10        -.21
[R.sup.2] (adjusted)                 .22          .19         .25

Note. N = 100-103.

(a) Levels of predictors in parentheses were those assigned higher
numerical values.

* p < .05. ** p < .01. *** p < .001.