Nasty women and bad hombres: the effect of racial and gender resentment on evaluations of presidential candidate valence

ABSTRACT

The competence, integrity, and experience of political candidates, i.e., their character valence, play an important role in voter decision-making. As character valence reflects the ability of a candidate to govern effectively and honestly – traits all voters generally value regardless of partisanship – candidates with higher valence should have an advantage in elections, yet 2016 saw candidates struggle to capitalize on their valence. Further, rhetoric in 2016 focused on the politics of resentment, the idea some groups were getting more than they deserve. We suspect the increased saliency of resentment in political campaigns affects valence evaluations of candidates, particularly when the candidates are a member of groups outside of power in our political system, including racial and ethnic minorities and women. Leveraging data from the American National Election Study, we find individuals with higher levels of resentment will more negatively assess the valence of candidates associated with the non-dominant groups, even among co-partisans, overriding their partisan tendency to view their candidate’s traits more favorably. These results suggest if resentment renders us unable to assess quality with any objectivity, candidates with less experience, acumen, and integrity may ride a wave of resentment sweeping more competent challengers away.

KEYWORDS:

• Candidate valence
• racial resentment
• gender resentment
• 2016 election
• race and ethnic politics

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 Theories of affective intelligence and emotional information processing potentially reflect the underlying mechanism behind our theory. First formalized by Marcus, Neuman, and Mackuen (2000), affective intelligence theory (AIT) reflects the idea that information processing switches between from a “dispositional system,” wherein people make decisions using habitual strategies when feeling positive emotions to a “surveillance system,” wherein voters update decision rules in the face of uncertainty and anxiety. Resentment may, in certain contexts, act as a trigger for anxiety, which in turn would have voters downplay valence characteristics, a common mode of analyzing individuals both within and outside the political sphere (Stone 2017), and instead look to other traits to make decisions. On the other hand, as shared resentment can trigger enthusiasm for a candidate, we might expect a stronger reliance on valence evaluation for the candidate that has turned an anxiety into a positive, active trait. That the motivating example of Marcus, Neuman, and MacKuen is the infamous Willie Horton ad from the 1984 US presidential election demonstrates the long-identified connection between racial resentment, fear, and candidate evaluations. Previous literature has demonstrated that everything from celebrity endorsements (Nownes 2017), to polling information can trigger shifts in emotional information processing, affecting not just what information voters use but how they act. As cross-pressured partisans are particularly apt to respond to this pressure reflects the impact of anger, which is itself triggered by the cognitive dissonance associated with conflicting gender resentment and partisan preference. The question becomes under what conditions voters switch from these two types of processing, i.e., when they reach their “affective tipping point” (Redlawsk, Civettini, and Emmerson 2010)? Our work highlights that resentment activation may be sufficient to activate anxiety as the primary emotional mode, especially for cross-pressured voters, yet in doing so, they do not become rational processors, but instead shift evaluatory metrics between different candidates. While it is outside the scope of this paper to specifically test the psychological mechanisms underpinning the relationship between resentment and valence discounting, we would like to thank an anonymous reviewer for suggesting this important connection.

2 Appendix Table A1 presents the summary statistics for the different components of the candidate valence measure. Each component is rescaled from 0 to 1, with scores closer to 1 indicating greater agreement with the question of whether the party’s candidate is a strong leader, really cares, is knowledgeable, honest, speaks their mind, and is even tempered.

3 Though there are differing levels of agreement across candidate evaluations in the valence components, modeling each component individually suggests evaluations of candidate characteristics are driven by the same factors. Appendix Tables A2 and A3 show consistency in the direction, significance, and to a lesser effect, magnitude, of the relationship between the covariates and the valence components in both the Democratic Models (Table A2) and Republican Models (Table A3). This approach suggests the additive nature of the valence scale is appropriate.

4 While racial and gender resentment are correlated, they are distinct, and appeal to distinct populations in the electorate. By combining them, we miss the distinctions that voters may draw between different outgroups. In addition, while resentment is impactful, we might expect that the impact of different forms of resentment are shaped by candidate characteristics or campaign saliency. By creating a single measure of resentment, we cannot examine their independent impact. The pre- and post-gender resentment measures are correlated at .45 and the pre- and post-racial resentment measures are correlated at .65. Racial and gender resentment are themselves correlated at .38, suggesting that resentment is rarely relegated to a single group, but that distinctions do exist.

5 We code independents as “0” indicating they are not Republicans for the purpose of a clear, binary measure. However, our results are robust to modeling the party measure as a 3-point measure, accounting for Democrats, Independents, and Republicans separately, with results indicating Democrats and Independents behave similarly with respect to the effect of racial and gender resentment.

6 Table A4 presents the summary statistics for the relevant variables.

7 The sample size for the base model is larger than for the resentment model, as the ANES data had fewer responses to the questions used to create the measures of resentment. Restricting the base model to only respondents in the resentment model produces results that are almost identical to those obtained using the full sample, and therefore we present the full sample for completeness. The summary statistics presented in Appendix Table A4 represent the summaries for respondents in both models.

8 Results are robust to modeling candidate valence separately. Appendix Table A5 presents separate models of Republican candidate valence and Democratic candidate valence. In the model of Republican candidate valence, party is conditioned by resentment, as it is in the differential models. For Democratic valence, however, the effect of party is not conditioned by resentment, with Republicans viewing the Democratic candidate less favorably, and both gender and racial resentment leading to less favorable evaluations of the Democratic candidate regardless of party.

9 Alternative specifications of the models considered if people of color or women responded uniquely to racial and gender resentment. A triple interaction between race, gender resentment, and racial resentment was not statistically significant. Additional alternative specifications accounted for interactions between party, race, gender, and the resentment measures. Consistent with the models presented in the paper, high levels of racial resentment were associated with increased favorability for the Republican candidate for Republicans, regardless of race, and for white Democrats. Non-white Democrats never favor the Republican, regardless of racial resentment. Similarly, high levels of gender resentment were associated with favorability of the Republican candidate for white Republicans. Levels of gender resentment above .9 were associated with favorability of the Republican candidate for non-White Republicans and white Democrats. Again, non-white Democrats never favor the Republican candidate, regardless of gender resentment. These results hold using a binary measure of race, as well as a four category measure accounting for individuals who identify as white, Black, Hispanic, and other non-white. The alternative specifications show our initial findings are robust – that resentment is a powerful force in evaluating candidate valence, affecting how even co-partisans evaluate candidates.

10 We include gender resentment as a potential influence for the 2012 election to demonstrate that, while resentment is a powerful political force in American politics, that not all forms of resentment are consistently salient in elections. This demonstrates a key part of our argument, that it is more than resentment, but resentment saliency, that leads to the effect on candidate valence evaluations.

11 The sample size in the 2012 models is larger than in the 2016 models due to the number of respondents in the ANES data.

12 The Chronbach’s alpha for this measure is lower than ideal, and therefore we replicated the models without the proxy measure for gender resentment and with alternative measures of gender resentment, and find comparable results offering evidence for the robustness of the models with the measure of gender resentment included.

13 Again, results are robust to modeling candidate valence separately. Appendix Table A6 presents separate models of Republican candidate valence and Democratic candidate valence. In the model of Republican candidate valence, party is not conditioned by resentment, as it is in the differential models. For Democratic valence, however, the effect of party is conditioned by resentment. While resentment makes members of both parties view the Democratic candidate less favorably, the effect is greater for Republicans, with evaluations of candidate valence dropping from .60 for Republicans with lowest levels of racial resentment to .32 at highest levels of racial resentment (a decrease of .28), compared to a .21 drop for Democrats over the same range of resentment (.74 to .53).