Abstract
It is known that the gender of candidates running for high levels of national office influences election outcomes more so than at lower levels, similar to the way in which women are advancing as middle managers but not as CEOs. Yet, the extent to which gender remains a barrier for candidates running for presidential office in contemporary American electoral politics is unknown. Understanding how gender bias, both conscious and unconscious, influences voting behavior is of crucial significance, as it has the potential to inform public policy by providing opportunities for rethinking and redefining the standards by which men and women are considered as presidential candidates by the American electorate.The 2016 presidential election can be treated as a natural case study to explore the possibility that the election outcome might have been influenced by sexism, both conscious (explicit) and unconscious (implicit) among voters. This study uses survey data to explore this possibility using measures of both explicit (self-report) and implicit (Brief Implicit Association Test) sexism. Inspiration for this study comes from a body of research that examined the extent to which racism, both implicit and explicit, influenced voter behavior in the 2008 election. While it is widely known that explicit sexism, such as that captured by the Hostile Sexism Inventory, contributes to a preference for male candidates over female candidates, less is known about the influence of implicit sexism, or sexism that a person may hold and be unaware of. It is believed that implicit biases, including gender bias, are natural, reflecting both evolutionary and cultural factors. However, implicit biases do impact how we evaluate one another and, subsequently, the decisions we make about who gets what in society.
The primary aim of this research is to explore if an implicit measure of sexism predicted vote choice for Donald Trump over Hillary Clinton in the 2016 election above and beyond the explicit measure. To compare implicit and explicit measures, this study added an abbreviated version of the Gender-Leadership Implicit Association Test to the second wave of the Cooperative Election Study (CES) that was fielded post-election in November 2017. Including an implicit measure of sexism is built on the premise that people might be unwilling to report sexist attitudes if asked to do so directly in a survey. Implicit measures do not rely on the honesty or accuracy of people’s answers to sensitive questions and, therefore, avoid social desirability bias.
As a secondary research aim, this study explores the potential that different types of sexism, such as hostile or benevolent influence vote choice. The premise for this analysis is based on Ambivalent Sexism Theory which posits that gender attitudes can be actively aggressive (hostile) or paternalistic and patronizing (benevolent). Ambivalent sexism thus reflects the duality of representations of cultural attitudes towards women. Based on this premise, we offer a framework for conceptualizing sexism in four primary ways that reflect an intersection of the implicit/explicit duality with that of the hostile/benevolent duality. Thus, sexism may be expressed as:1) explicit benevolent sexism; 2) implicit benevolent sexism; 3) explicit hostile sexism; and 4) implicit hostile sexism. The analysis generates a critique of the Brief Implicit Association Test (BIAT) regarding its effectiveness in capturing implicit sexism across various contextual and situational factors.
The original design of the study included one measure of explicit sexism, which was an abbreviated version of the Hostile Sexism Inventory, with which to compare the implicit measure. However, informed by the theory of Ambivalent Sexism, the Gender Equality Scale was added as an ad hoc attempt to capture explicit benevolent sexism since an abbreviated version of the Benevolent Sexism Inventory was not included on the original survey. Adding the Gender Equality Scale as a measure of explicit benevolent sexism allowed for analysis of the intersection of implicit and explicit measures with type of sexism. This was particularly useful since the current Gender-Leadership Implicit Association Test relies primarily on the construct of benevolent sexism while the primary explicit measure relies on the construct of hostile sexism. Results of this secondary analysis are informative for the future development of implicit measurement instrumentation, the nature of implicit bias, and interventions to reduce bias.
The theoretical frameworks for this research are multifaceted and draw from various disciplines including sociology, political science, and psychology. First, the concept of implicit bias is introduced using broad conceptual frameworks that explain the origins of social biases and theorize intergroup behavior from a sociological perspective. These theories include Realistic Group Conflict (Campbell 1906-1988), Integrated Threat Theory (Stephan, 2000), and Social-Identity Theory (Tajfel, 1971). Drawing on these broad frames, Feldman’s theory of inference in political perception is used to disentangle important interpersonal differences in how voters perceive the image (i.e. race, gender) of a candidate. Finally, Ambivalent Sexism Theory (Glick and Fiske, 1996) is used to inform social perception of women in the context of a national presidential election. Research hypotheses were developed using the theory of implicit bias and Ambivalent Sexism to predict voter behavior in the 2016 election with the possible outcome of voting for Trump or voting for Clinton.
The results of this dissertation contribute to a deeper understanding of the conscious and unconscious sources of sexism that influence voter behavior in modern American politics and develop a better understanding of the concept of implicit bias in general and the methodological features of implicit cognition measurement. The implications of these findings have the potential to inform a range of policies, from the way women run their political campaigns to increasing the efficacy of diversity trainings and improving methodology for capturing implicit biases.