Women Voters and the Utility of Campaigning as “Women of Color”

ABSTRACT

Women candidates from underrepresented racial and ethnic groups must cast a wide net when attracting voter support to win elections. One aspect of their candidate profile is the ability to campaign explicitly as a “woman of color.” Despite increased media usage of the term women of color, few academic studies focus on whether and how this term resonates with voters. We analyze original survey data collected in 2020 to probe voter reaction to the women of color identity among self-identified Black women, Latinas, and white women. We evaluate whether women are aware of the term and analyze the importance they place on the election of women of color candidates. Finally, to understand the effectiveness in actual electoral contests, we conduct an experiment to determine if women’s evaluations of Kamala Harris are affected by a woman of color candidate frame.

KEYWORDS:

• gender
• race
• elections
• intersectionality
• identity

Acknowledgments

The authors thank the editors and anonymous reviewers for their helpful comments.

Disclosure statement

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

Supplementary material

Supplemental data for this article can be accessed on the publisher’s website at https://doi.org/10.1080/1554477X.2022.2007467

Notes

1. Upon becoming vice-president elect, Harris noted: “While I may be the first woman in this office, I will not be the last.” https://www.theguardian.com/us-news/2020/nov/07/kamala-harris-victory-speech-first-woman-vice-president

2. CAWP Reports – on Gender and Elections (https://womenrun.rutgers.edu/)

3. Representatives Alexandria Ocasio-Cortez of New York, Ilhan Omar of Minnesota, Ayanna Pressley of Massachusetts, and Rashida Tlaib of Michigan.

4. President Trump tweeted: “So interesting to see ‘Progressive’ Democrat Congresswomen, who originally came from countries whose governments are a complete and total catastrophe, the worst, most corrupt and inept anywhere in the world (if they even have a functioning government at all), now loudly and viciously telling the people of the United States, the greatest and most powerful Nation on earth, how our government is to be run.” https://www.washingtonpost.com/politics/trump-says-four-liberal-congresswomen-should-go-back-to-the-crime-infested-places-from-which-they-came/2019/07/14/b8bf140e-a638-11e9-a3a6-ab670962db05_story.html; https://www.washingtonpost.com/politics/trump-calls-on-minority-congresswomen-to-apologize-after-he-said-they-should-go-back-to-their-countries/2019/07/15/897f1dd0-a6ef-11e9-a3a6-ab670962db05_story.html

5. For exceptions see Sanbonmatsu (2020) and Matos, Greene, and Sanbonmatsu (2021).

6. The 2016 Collaborative Multiracial Post-Election Survey (CMPS) included questions about candidates identified by different racial/ethnic and gender categories but not by the label WOC. Lemi (2020a) conducted an experiment that manipulated information about Harris’s multiracial background but did not consider the WOC label.

7. Although it was not planned, the first day of our field period was the day after Kamala Harris accepted the nomination for vice president at the Democratic National Convention. The survey was conducted between August 20 and August 26, 2020; the median completion time for Black women, Latinas, and white women was 9 minutes. The IRB designated the study exempt from full IRB review. Table A1 in the Appendix contains descriptive statistics and compares our data with CPS and CCES 2019 benchmarks. These statistics reveal that our data resembles nationally representative samples and so we do not weight the data in this paper. However, we caution readers that the Black women and Latinas in our study are younger than Black women and Latinas in the population. Compared with the population, the Latinas in our sample who are over 50 are less well-represented than older Black women. The difference between weighted and unweighted means on our main dependent variables of interest are negligible. We selected Qualtrics in order to survey an adequate number of Black and Latina respondents. Evidence comparing nonprobability and probability vendors point us to the credibility of some nonprobability internet samples (Kennedy et al. 2016). Qualtrics has also been found to outperform other nonprobability samples in comparison to GSS (Zack, Kennedy, and Long 2019).

8. Because of resource limitations, we were not able to include sizable numbers of women from all racial/ethnic backgrounds. Future research can extend our investigation and assess how the WOC candidate label affects women who identify as Asian American, Native American, and MENA. Throughout the paper, we use “white women” to refer to non-Hispanic white women, recognizing that Latinas can be of any race.

9. For those who chose both “Black” and “Latina” on the race question, we randomly assigned half to the Black sample and half to the Latina sample. Race/ethnicity was measured with the question: “Which of the following racial or ethnic group(s) best describes you? [Check all that apply]”. We exclude 27 white women who checked more than one racial group.

10. We did not provide a definition of WOC, leaving it to the respondent to interpret the identity.

11. Mean responses on the five-point scale about the importance of candidate being WOC (with 5 representing “extremely important” and 1 “not important at all) were 3.76 (Black women), 3.37 (Latinas), and 2.36 (white women).

12. In the Matos, Greene, and Sanbonmatsu (2021) study, the number of Latinas included in the 2018 CCES dataset was limited, though Latinas appeared to fall in between Black and white women in support for electing more WOC to Congress.

13. The order of the two questions was randomized.

14. Mean responses on the five-point scale about the importance of a candidate being a woman (with 5 representing “extremely important” and 1 “not important at all) were 3.80 (Black women), 3.59 (Latinas), and 2.70 (white women). These differences in how each group of women rated WOC candidates versus women candidates are statistically significant.

15. We limit the analysis to those respondents familiar with the WOC term. Controlling for question order – whether the WOC candidate identity or woman candidate identity question appeared first – does not change our multivariate results.

16. See Orey (2004).

17. By using racial resentment, rather than linked fate, for example, we are interested in whether expressions of racial resentment hinder some women’s ability to support a WOC candidate.

18. Education is a 7-point scale; political interest is a 5-point scale; and voted is a dichotomous variable.

19. The two screener questions used closed-ended questions to ask respondents about the topic of the Qualtrics survey. We categorize “high attention” respondents as those who answered both screener questions correctly, and “low attention” respondents as those who answered neither question or one question correctly. Consistent with Berinsky, Margolis, and Sances (2014), we view survey attentiveness as a latent construct and retain low-attention respondents in the paper. We are aware of the limitations of nonprobability samples but we believe our data offer unique and important evidence of the WOC candidate label and its meaning for different racial groups of women. See the Online Appendix for versions of Tables 1–4 that provide responses by survey attentiveness.

20. Berinsky, Margolis, and Sances (2014,740) conclude that “presenting results stratified by levels of attentiveness” is one strategy to enhance research transparency in studies relying on self-administered surveys.

21. Partisanship is coded 1 = Republican, 2 = Independent, and 3 = Democrat.

22. Afro-Latinas are also younger than other Latinas in our sample.

23. We also estimated a model using ideology rather than partisanship since the two variables are correlated. Ideology behaved similarly to partisanship in this alternative specification (full results in the Online Appendix).

24. At the time of our study, Harris was a U.S. Senator and had been nominated to be the vice presidential candidate for the Democratic party.

25. We did not include a condition that manipulated Harris’s identity as both a Black woman and South Asian woman, although future research could examine a multiracial frame compared with an explicit WOC frame. Past work by Lemi (2020a) examined Harris’s multiracial identity in her 2016 U.S. Senate race, finding that support declined for Blacks who reported strong racial identities when Harris was identified as Black and South Asian compared to the absence of any racial information about Harris or her opponent, Loretta Sanchez.

26. In addition, the South Asian woman frame lowered Harris’s evaluation by 0.33 compared with the Black woman frame (p = .11) for low-attention Latinas, although this result is not statistically significant.

27. Full results are available in the Online Appendix.

28. Clayton, Crabtree, and Horiuchi (2020) investigated the “woman,” “Black woman,” “Asian woman,” and “Black and Asian woman” frames, as well as a condition in which no gender or race was included.

Additional information

Notes on contributors

Stacey Greene

Stacey Greene is Assistant Professor of Political Science at Rutgers University, New Brunswick.

Yalidy Matos

Yalidy Matos is Assistant Professor of Political Science and Latino and Caribbean Studies at Rutgers University, New Brunswick.

Kira Sanbonmatsu

Kira Sanbonmatsu is Senior Scholar, Center for American Women and Politics, Eagleton Institute of Politics, and Professor of Political Science at Rutgers University, New Brunswick.