The Structure of Presidential Evaluations: White Men, White Women, and Trump

ABSTRACT

This study uses an intersectional lens to study the structure of presidential evaluations among white men and white women. I argue that due to their distinct positions of power in American politics via the intersection of race and gender, different processes, wrapped into white heteropatriarchy, explain white men and women’s 2018 favorable mid-term evaluations of President Donald Trump. Using a sample from the 2018 CCES, I find that although there is some overlap in the factors that explain positive evaluations of Trump anti-Blackness, gender identity, and Republican partisanship structured white men’s favorable evaluations of Trump, while sexism, anti-immigration sentiment, and ideology structured white women’s favorable evaluations. The variable with the strongest association with evaluations of Trump for men was negative assessments of the economy, and the variable with the strongest association with evaluations of Trump for women was anti-immigration sentiment.

Keywords:

• gender
• race
• white identity
• presidential evaluations

Disclosure statement

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

Notes

1. I thank a colleague for this insight.

2. https://www.cnbc.com/video/2018/10/01/trump-were-getting-another-1-point-6-billion-for-the-border-wall-this-year-border-security-immigration-mexico-canada-nafta-trump-trade-deal-trade- war.html. Comments start at 1:15 mark.

3. I thank a reviewer for suggesting this. Given the assumptions needed to test mechanisms (e.g., Imai et al. 2011), I do not directly test partisanship as a mechanism and instead treat Republican identification as a control variable.

4. I thank a reviewer for making the suggestion to apply Young (2003) to this hypothesis.

5. https://www.huffpost.com/entry/it-wasnt-just-white-men-who-participated-in-the-unite-the-right-rally_n_598f55b4e4b09071f69a0381?ncid=engmodushpmg00000006

6. 2.03% of the sample said “not sure.”

7. For importance of racial and gender identity, 6.62% said don’t know for race, 3.38% said don’t know for gender. Nearly 30% of the CCES white sample said don’t know for race linked fate, and nearly 21% of the white sample said don’t know for gender linked fate.

8. The results in Table 2 change slightly when I treat the outcome variable as continuous, where 1 = strongly approve, 4 = strongly disapprove, and “not sure” is treated as missing. Gender identity and age lose statistical significance for men, and Republican partisanship becomes statistically significant for women. The signs also flip, but that is likely because the outcome variable is coded differently. Given the non-continuous nature of the outcome variable, I am cautious about what this means substantively, but this alternative set of findings indicates that for the most part, the results in Table 2 are robust to modeling the outcome variable with OLS or logit (see Appendix B). When I expand the linked fate variables and hold “no” as the baseline, expand the racial and gender identity variables holding “not at all important” as the baseline, and expand the employment status variable holding “full-time” as the baseline, the results for men change slightly. For men, gender identity and anti-immigration sentiment lose statistical significance. For women, age loses statistical significance and employment status becomes statistically significant. These results should also be treated with caution since the model for men noted 4 instances of perfect collinearity and the model for women noted 16 instances of perfect collinearity (Appendix C). When I exclude those who said “don’t know” to the gender linked fate question by treating those responses as missing, the results produce a collinearity problem for men (1 failure and 0 successes completely determined), and Republican partisanship and anti-immigration sentiment lose statistical significance. For women, the original model retains statistical significance (Appendix D). I interpret these results as evidence that most of the original model, with its original coding, remains robust to minor changes in variable specification.

9. Plots created with Bischof (2017).

Additional information

Notes on contributors

Danielle Casarez Lemi

Danielle Casarez Lemi is a Tower Center Fellow at Southern Methodist University.