2016 Election Reflection Series: Women and Support for Trump

Prior to the 2016 election, APSA’s Diversity and Inclusion Programs Office issued a call for scholarly reflections, original research notes, and classroom exercises that shed light upon diversity, political behavior, public opinion and the 2016 Campaign and Election. What resulted is an eight part series, 2016 Election Reflections, covering a range of election related topics and research methods.    


Women and Support for Trump: Race, Place, Identity or Something More?

by Gregory Davis, Harvard University

With Donald Trump’s victory in the U.S. presidential election, many political commentators and social scientists have tried to figure out where the Republican Party picked up the votes necessary to reverse their fortunes after losing in 2012. One demographic that is of particular interest is women. Hillary Clinton, the first female presidential candidate of a major political party, did worse with women than Barack Obama did in 2012, 54% to 55% (CNN). What’s more, Donald Trump, repeatedly dogged as an explicit sexist in the campaign, beat Clinton among White women by a 10-point margin (43% to 53%). Needless to say, this result raises many questions.

For one, what motivated different electoral outcomes for women in this election? There are a few sources we can point to. One is demographics. Beyond simple party affiliation, race, religion, education, income, residential area, SES, and age may have exerted influence on women to pick Trump over Clinton. In addition, strength of identity – as a woman, as a member of one’s racial group, or in terms of religious identity – may have independently impacted women’s choice of candidate. Last, I, along with Professor Jim Sidanius and colleagues Kiera Hudson and Asma Ghani at Harvard University, were interested in the added impact of worldview variables such as Social Dominance Orientation (SDO, preference for group-based social and power hierarchy; Sidanius, 1999) and Right-Wing Authoritarianism (RWA, preference for a traditional and conformist world order lead by a strong authority; Altemeyer, 1998) in how women chose candidates. We found that while demographics did predict some of the variance in candidate choice and candidate support, the strength of identity variables did not. Also, the aforementioned worldview variables SDO and RWA independently and strongly predicted these outcomes, even after controlling for party affiliation. Strength of identity and worldview measures like SDO and RWA are rarely studied in predicting candidate support, yet the data discussed here highlight their great importance or lack thereof.

Our Sample and Our Measures
In October 2016, a month before the presidential election, we conducted a large-scale panel survey of likely voters in the United States online via Amazon Mechanical Turk. That sample included 1,587 women who were 76% White, 51% suburban, and 54% college educated. The median ages were 23 and 50 years and the sample represented the full scale of income levels and religious affiliations. We asked these women to choose their preferred candidate, between Clinton and Trump, and to then answer a seven-question scale about participants’ opinions of their chosen candidate and if they thought he or she would make a good president, which we called the Candidate Support Scale (ɑ = .94, mean = 5.56/7).

Next we asked these women for scale values of the strengths of their gender identity (Luhtanen & Crocker, 2000; ɑ = .86, mean = 4.86/7), racial identity (Sidanius et al., 2000; ɑ = .85, mean = 3.71/7), and religiosity (ɑ = .95, mean = 2.25/5). These scales included questions like, “In general, my racial group is an important part of my self-image,” “Being a man/woman is an important reflection of who I am,” and “My religion is central to my identity,” respectively. Last, we measured social psychological worldview scales such as SDO (ɑ = .92, mean = 2.50/7) and RWA (ɑ = .95, mean = 2.70/6). SDO includes statements like, “Some groups of people must be kept in their place,” and, “We shouldn’t try to guarantee that every group has the same quality of life.” RWA, on the other hand, prompts participants to agree or disagree with statements like, “Our country desperately needs a mighty leader who will do what has to be done to destroy the radical new ways and sinfulness that are ruining us,” and, “This country would work a lot better if certain groups of troublemakers would just shut up and accept their group’s traditional place in society.”

Predicting Women’s Choice of Trump over Clinton
After analysis, we see that all three of our sources – demographic groups, feelings of identity, and worldview variables – significantly predicted the likelihood of preferring Trump to Clinton in the election even after controlling for party affiliation. Compared to White women, Black women in our sample were about 60% less likely to pick Trump instead of Clinton, a result that nonetheless loss significance as variables were added to the model. Conversely, Jewish women were 84% less likely to pick Trump as evangelical women, a difference that remained significant throughout. As a counter, women with less than a high school diploma were more than five times more likely to pick Trump than were women with a bachelor’s degree. What’s more, each year of age brought on a 2% increase in the same likelihood. There were no effects seen for income, subjective SES, or residential area (urban vs. rural vs. suburban).

As remarkable as many of these effects were, they combined to explain only 6.1% more of the variance among the women’s choices than party affiliation alone. Once we added our strength of identity variables (racial, gender, and religious), we only explain another 0.2%. Importantly, there were no added effects for any of these variables. It’s particularly interesting that gender identity had no effects in the model, as it indicates that the association between womanhood and preferring Hillary Clinton was nil for the women in our study. This nonsignificant effect for strength of gender identity, moreover, still exists if we don’t control for party affiliation.

Last, we added SDO and RWA to the model. This model explained 52.9% of the variance (4.0% more than without these worldview variables) and the variables were each greatly predictive of women supporting Trump over Clinton. Each unit increase in SDO predicted an increased likelihood of picking Trump by 44%. RWA was even more impactful, predicting an increased likelihood of 74% per unit.

Predicting the Strength of Support for Women Trump Supporters
Isolating the 526 women who ultimately chose Trump, we used the same variables to predict their scores on the Candidate Support Scale. Other than party affiliation, the only demographic category that was significantly predictive was race, where Black women were less supportive of Trump’s candidacy than White women, scoring on average 4.1 points on a 1-to-7 scale of support compared to White women’s 5.6. Age and religion played a role as well, with older and Catholic Trump-supporting women having stronger Candidate Support Scale scores than younger and Protestant female supporters, respectively.

After adding strength of racial identity, gender identity, and religiosity to the model, these variables had no effect on predicting support for Trump for these women. Finally, SDO failed to have any impact on how much these women supported Donald Trump. RWA, however, predicted an increase in support for Trump, by nearly half-a-point each unit above the mean. The final model using the demographic information, the identity variables, SDO, and RWA, predicted 37.8% of the variance in the Candidate Support Score measure for Trump-supporting women.

Implications and Concluding Remarks
By surveying a large group of female Trump and Clinton supporters, we were able to shed considerable light on the social psychological influences behind choosing either of these unique candidates. While we found some persistent demographic differences in candidate choice and support, psychological variables like Social Dominance Orientation and Right Wing Authoritarianism explained a large amount of the variance. RWA – a measure of support for strong authoritarians to maintain order and conformity by force – was a particularly strong predictor of both which women chose Donald Trump over Hillary Clinton, and also how much they supported him.

Social psychologists frequently use SDO and RWA to study outcomes like outgroup hate, perceptions of threat, and willingness to oppress low-power groups. Here, we used these variables to predict preference for a presidential candidate and support of their candidacy. We found that they strongly and significantly predict these outcomes even controlling for factors like demographics and party affiliation. Although SDO and RWA are highly correlated with an identity as a conservative, subsequent analysis by our lab showed that SDO and RWA are still independent predictors of choosing Trump over Clinton even when controlling for conservative identity in addition to party affiliation. Thus, Social Dominance Orientation and Right Wing Authoritarianism display unique and independent effects in the realm of candidate support, and require further study.

As a critical note about the data analysis reported here, the strength of support scale measures – for racial identity, religious identity, and most importantly gender identity – did not predict anything when entered into our models. This sets up important implications that require further research. If, for example, strongly identified women are no more or less likely than weakly-identified women to choose a female candidate for president over a male one, then that suggests that the identity salience of being a woman is not necessarily tied to showing solidarity with other women, at least not in the realm of American politics. Drilling down further, our research suggests that highly-identified women who choose the male candidate are no different in the strength of their support than lowly-identified women despite the fact that the male candidate in question (President-Elect Donald Trump) has a well-publicized history of making degrading remarks about women.

Importantly, these data were gathered from October 4, to October 6, 2016, and the Access Hollywood video revealing Donald Trump’s vulgar comments about a woman were publically released October 7, 2016. It is certainly possible that had we collected these data a week later than when we did, we would have found significant results with the strength of gender identity measure. Nevertheless, we believe that these results regarding strength of identity are not just the result of timing, as the President-Elect’s prior remarks about women were highlighted heavily in the first presidential debate, which occurred more than a week before we collected these data, on September 26, 2016.

As a whole, these data can shed some light on how women decided which candidate appealed to them, and which women went in either direction. Women’s choice between Hillary Clinton and Donald Trump was not a simple one, and aspects of a woman’s life like party identification, demographics, strength of identity, and her worldviews about dominance and authoritarianism each provide only part of the answer. Hopefully, we can use this knowledge to expand political science and social psychology literature on how ideology, identity, and political choice interact beyond what demographics alone can predict.

Gregory Davis is a doctoral student in African American Studies at Harvard University. He studies the impact of personal worldviews on how we evaluate and respond to minority targets and policies that may benefit them.


Altemeyer, B. (1998). The “other” authoritarian. Advances in Experimental Social Psychology, 30, 47-92.

Luhtanen, R., & Crocker, J. (1992). A collective self-esteem scale: Self-evaluation of one’s social identity. Personality and Social Psychology Bulletin, 18, 302-318.

Sidanius , J. , Levin, S., Liu, J.H. & Pratto, F. (2000). Social dominance orientation and the political psychology of gender: An extension and cross-cultural replication. European Journal of Social Psychology, 30, 41-67.

Sidanius, J. & Pratto, F. (1999). Social dominance: An intergroup theory of social hierarchy and oppression. Cambridge, UK: Cambridge University Press.