Countering Implicit Appeals: Which Strategies Work?

Abstract

Some contemporary politicians try to mobilize racial attitudes by conveying implicit racial messages against their opponents—messages in which the racial reference is subtle but recognizable and which attack the opponent for alleged misdeeds. Although targeted politicians have tried a number of different strategies to respond to implicit racial appeals, little is known about the effectiveness of these strategies. Using two survey experiments, we answer the following question: Does calling the appeal “racial” work? That is, does it neutralize the negative effects on the attacked candidate? We find mixed evidence that it does. However, offering a credible justification for the attacked behavior works more consistently. We also test whether effects vary by candidate race. The results suggest that Black candidates’ rhetorical strategies are more constrained than identical White candidates’, but that White Americans are more open to credible arguments and justifications than the previous literature implies.

Keywords:

• race
• opinion

Notes

1. OkieCampaigns’s channel (2008, September 5). Detroit mayor Kwame Kilpatrick / Obama ad [Video file]. Retrieved from http://www.youtube.com/watch?v=iNyZ-cbPdoM

2. Mooney, M. (2008, August 1). Obama aide concedes “dollar bill” remark referred to his race. ABC News. Retrieved from http://abcnews.go.com/GMA/Politics/story?id=5495348&page=2

3. Mimic is the strategy of signaling to voters that one is no less sympathetic to White voters’ views than the other party, as Bill Clinton did when he condemned the musician Sister Souljah for her statements about the Los Angeles riots during the 1992 campaign. We set it aside because mimic is used less as a way of defending oneself against attack (which is the scenario explored here) and more as a way of going on the offensive (as Clinton did in his Sister Souljah comments).

4. Gewargis, N. (2008, July 31). McCain camp: Obama “playing the race card from the bottom of the deck” [Web log post]. Retrieved from http://abcnews.go.com/blogs/politics/2008/07/mccain-camp-oba/

5. Rosenthal, A. (1988, October 24). Foes accuse Bush campaign of inflaming racial tension. The New York Times. Retrieved from http://www.nytimes.com/1988/10/24/us/foes-accuse-bush-campaign-of-inflaming-racial-tension.html? pagewanted=all&src=pm

6. Shear, M. S. (2011, April 27). Obama releases long-form birth certificate [Web log post]. Retrieved from http://thecaucus.blogs.nytimes.com/2011/04/27/obamas-long-form-birth-certificate-released/

7. We do not label the engage-nonracial as an implicitly racial rebuttal because in a White-White contest, it acts as a nonracial rebuttal.

8. Toner, R. (1988, October 20). Dukakis makes strong response to G.O.P.’s ads. The New York Times. Retrieved from http://www.nytimes.com/1988/10/20/us/dukakis-makes-strong-response-to-gop-s-ads.html

9. To be sure, other studies find that factual information produces little opinion change and in some cases leads to a backlash effect. The topics of such studies include the Iraq War and its rationales (Berinsky, 2007; Nyhan & Reifler, 2010) and immigration policy (Sides & Citrin, 2007). The null effects of factual information usually involve issues that the public already has firm (albeit uninformed) opinions about. In contrast, the significant effect of factual information occurs with lesser-known issues such as foreign aid and school spending. We expect that the campaign setting in this study more closely resembles a lesser-known issue: most voters know very little about state-level elected officials (Delli Carpini & Keeter, 1996). Therefore, credible information may influence opinion about politicians that voters know little about.

10. In other words, we assign Democratic respondents to a Democratic primary and Republican respondents to a Republican primary. Independent respondents are randomly assigned to either primary.

11. Questions appear in the order that they are described here. Complete question wording and ordering is included in the Question Wording section of the Supplemental Material.

12. This may be why John McCain did not make more use of the tactic in 2008 (Tesler & Sears, 2010, p. 55).

13. The faces used in the preliminary study are shown in Figure 1 of the Supplemental Material.

14. Wells is the sitting governor of the state, which explains how he could have issued a pardon.

15. Although Kilpatrick was arrested for assaulting a police officer, we added the detail about the officer requiring hospitalization. Our aim was to make the assault seem severe enough to capture respondents’ attention in an Internet interview mode where that attention is scarce.

16. Since the target and criminal are manipulated to be either Black or White, the four possible combinations are Black target/Black criminal, White target/Black criminal, Black target/White criminal, and White target/White criminal.

17. Notice that no engage-racial or engage-justify + racial rebuttal is given in the White target/White criminal scenario because it would not make sense for a White target to offer a racial rebuttal when tied to a White criminal.

18. If we did not paint the White baseline face, morphing the White baseline face with the Black mixing face would yield a racially ambiguous face. It is important that the faces are clearly recognizable as Black or White in the experiment.

19. According to Oosterhof and Todorov (2008), dominance and trustworthiness account for 80% of social judgments, so there is little utility in measuring all nine dimensions.

20. See Supplemental Material section 9.8 for full text of each rebuttal condition.

21. Sterling, A. (2009, December 1). Huckabee calls criticisms over clemency “disgusting.” CBS News. Retrieved from http://www.cbsnews.com/8301-503544_162-5854298-503544.html

22. Twenty-eight (3.7%) of the 766 subjects in the main study were not recruited from MTurk, but instead were recruited from the Princeton Survey Research Center’s Mercer County (NJ) panel.

23. As a further check that the message functioned as a negative racial message, we examined and found a racial effect among subjects who answered all manipulation check questions correctly and whose resentment score fell in the top third of the sample (above 0.625).

24. Covariates are similar across the rebuttal conditions, verifying that the groups are equivalent before the treatment, thus meeting the ignorability assumption (Table 3 in the Supplemental Material).

25. For the contrast with White target/White criminal, each rebuttal is considered except those without a White target/White criminal condition, namely, engage-racial and engage-justify + racial, which are not realistic rebuttals for a White target paired with a White criminal.

26. Target/criminal pairs will be described using the target’s race first and the criminal’s race second. For example, “White/Black” refers to a White target associated with a Black criminal.

27. Put differently, the FDR adjustment controls for “the number of false discoveries as a proportion of the number of true discoveries” (Newson, n.d., p. 10).

28. The adjusted alpha level accounts for the 103 tests we conducted (22 comparisons to 0; 30 within-rebuttal comparisons [e.g., Black/Black engage-racial versus White/Black engage-racial]; and 51 within-pair comparisons [e.g., Black/Black engage-racial versus Black/Black engage-nonracial). For calculation details, see Appendix B of Larson-Hall (2009).

29. Confidence intervals are set at 98% (instead of 99%) for tests relative to 0 or to ignore, because predictions are directional (i.e., rebuttal effects are positive) and thus significance tests are one-tailed. We expect positive (and thus unidirectional) effects only relative to 0 and ignore (criteria 1 and 2), but not for within-rebuttal or within-pair comparisons (criteria 3, 4, 5, and 6).

30. As a simple illustration, imagine a respondent who dislikes the target and likes the attacker following the attack, but approves of the target’s rebuttal and adjusts his post-rebuttal evaluation of the target upward and the attacker downward. If the respondent gave pre-rebuttal scores of 30 to the target and 50 to the attacker and post-rebuttal scores of 40 to the target and 35 to the attacker, the post-pre change in the difference score for that respondent would be 40 - 35 (which is the post-rebuttal difference) minus 30 - 50 (the pre-rebuttal difference) = 25—a strong endorsement of the target’s rebuttal.

31. The only exception is White/Black distract, which is significantly more effective than 0 (p < 0.01, one-tailed), but is not significantly more effective compared to ignore (p = 0.02, one-tailed).

32. A test of each pair to the White target/White criminal pair (criterion 3) is not possible because we did not consider engage-racial or engage justify + racial to be a realistic strategy for a White target tied to a White criminal. Also, significance tests are two-tailed since we do not have strong expectations that a White target’s rebuttal will be more effective than a Black target’s rebuttal.

33. The smallest p-values for any of these tests was 0.1 when comparing Black/Black versus White/White.

34. The smallest p-value for criterion 3 tests (White/White versus the other three pairs) was 0.06 for White/White versus Black/White.

35. The smallest p-value is 0.89 for White/Black versus Black/White comparison.

36. It is marginally more effective for the White/Black pair compared to White/White (p = 0.03). Black/Black and Black/White are no less effective compared to White/White (p = 0.85 and p = 0.76, respectively).

37. Two-tailed tests are used for ignore since it is a non-rebuttal and therefore is not expected to improve evaluation of the target.

38. Ignore is marginally less effective for White/White compared to Black/Black and to White/Black (p = 0.07 for each comparison, two-tailed test), indicating that when paired with a Black criminal, a target of any race may find it more beneficial to stay silent than he would if race was not a factor in the campaign. However, the marginal significance and lack of strong theory prevents us from drawing clear conclusions on this.

39. Figure 5 in the Supplemental Material shows the effect of the rebuttals on the pre-/post-rebuttal change in the target’s feeling thermometer score. These results are similar to the feeling thermometer difference score analysis shown in Figure 3, which takes into account feeling thermometer ratings of the attacker. There are no significant differences at the 0.05 level between the difference score estimates and the target feeling thermometer estimates. Also in the Supplemental Material are Figures 6, 7, 8, and 9, which show the effects of the rebuttals on approval of the rebuttal and ratings of the target’s trustworthiness, competence, and ideology. The results for these ratings look similar to the difference score estimates though they are weaker. Figure 10 in the Supplemental Material shows the effect of the rebuttals on voter turnout intention. None of the rebuttals appear to have a mobilizing effect.

40. The smallest p-value is 0.14 when comparing engage-racial to engage-factual + racial.

41. As noted earlier, all but distract are more effective than ignore (see Figure 4 in the Supplemental Material).

42. We use 94% confidence intervals because they correspond with a two-tailed test with alpha = 0.01 when examining overlap of their confidence intervals (MacGregor-Fors & Payton, 2013).

43. The adjusted alpha level accounts for the 82 tests we conducted (19 comparisons to 0; 27 within-rebuttal comparisons; and 36 within-race-pair comparisons).

44. Figure 12 in the Supplemental Material shows the effect of rebuttals compared to ignore for the preliminary study.

45. One key difference between the two studies that may explain this is the improved realism of the faces used in the main study (see Figures 2 and 3 in the Supplemental Material) compared to those used in the preliminary study (Figure 1 in the Supplemental Material).

46. The exact wording of the question was, “Think about Michael Wells, the governor who pardoned Jones. In your opinion, does the phrase, ‘he is competent’ describe Michael Wells extremely well, quite well, not too well, or not well at all?” The answer choices were 1 = extremely well, 2 = quite well, 3 = not too well, and 4 = not well at all. We reverse coded the answer choices so that 1 corresponds to low competence and 4 to high competence.

47. The exact question wording was, “Do you think of Michael Wells, the governor who pardoned Jones, as extremely liberal, liberal, slightly liberal, moderate or middle of the road, slightly conservative, or extremely conservative?” The answer choices were 1 = extremely liberal, 2 = liberal, 3 = slightly liberal, 4 = moderate/middle of the road, 5 = slightly conservative, 6 = conservative, 7 = extremely conservative.

48. When we look only at those who are “very interested” or “somewhat interested” in politics and public affairs, the results do not change. This takes into account that some degree of political sophistication is necessary to understand the implications of ideological labels such as liberal or conservative.

49. Haberman, M. (2012, August 7). Bill Clinton slams Romney “misleading” welfare ad [Web log post]. Retrieved from http://www.politico.com/blogs/burns-haberman/2012/08/bill-clinton-slams-romney-misleading-welfare-ad-131405.html

Additional information

Notes on contributors

Matthew Tokeshi

Matthew Tokeshi is a Ph.D. candidate in the Department of Politics, Princeton University. Tali Mendelberg is Professor of Politics, Department of Politics, Princeton University.

Tali Mendelberg

Matthew Tokeshi is a Ph.D. candidate in the Department of Politics, Princeton University. Tali Mendelberg is Professor of Politics, Department of Politics, Princeton University.