Who stereotypes female candidates? Identifying individual differences in feminine stereotype reliance

AbstractIncreasing numbers of women are running for political office at the local, state, and national levels. Existing research offers unclear conclusions about whether feminine stereotypes are an electoral constraint for female candidates. An underlying assumption in this scholarship is that all types of individuals rely on similar processes to form electoral assessments of female candidates. This study tests the assumption of equitable stereotype reliance across individuals. I integrate theories from psychology about which types of individuals are most likely to use stereotypes to judge others, and consider how these determinants operate in a political context. I argue that whether an individual relies on feminine stereotypes to evaluate a female candidate depends on characteristics such as attention to politics, partisanship, and other relevant demographic characteristics. An original survey experiment identifies how individual characteristics affect whether a voter turns to feminine stereotypes when a woman runs for office. These findings are consequential because individuals who rely on feminine stereotypes are also less likely to vote for a female candidate.Keywords: gender stereotypesvote choicefemale candidatesrepresentation Notes1. For recent studies considering the role of context and information in increasing stereotype saliency for female candidates, see Krupnikov and Bauer (Citation2014) and Bos (Citation2011).2. I refrain from using photos to cue candidate gender as the appearance of the candidates may affect stereotyping (Rosenberg et al. Citation1991). These names were pre-tested with a convenience sample from Amazon's Mechanical Turk (N = 129). Participants in the pre-test rated hypothetical individuals named Susan Foster and Tom Larson as equitable in terms of their age (p = 0.1460), perceived education level (p = 0.9887), warmth (p = 0.3640), and emotionality (p = 0.3881).3. Partisanship was asked before the stimulus so that participants could be appropriately sorted into partisan conditions. I measure partisanship with a 7-point scale with the categories ranging from being a strong partisan, partisan, weak partisan, and Independent/non-partisan. To sort participants into the appropriate conditions, those who identified as Independent/non-partisan answered a follow-up question asking which party they leaned most closely toward. Participants identifying as Independent/non-partisan answered a second question about the party they leaned more closely toward. All participants who initially identified as Independent/non-partisan (27%) selected a preferred party.4. There is some debate in the literature about the utility of news attention given the tendency of participants to overestimate how frequently they follow the news (Price and Zaller Citation1993; Prior Citation2009; Dilliplane, Goldman, and Mutz Citation2013). The mean value for the measure is 3.04 (SD = 1.45), and this is in line with the comparable items in the 2012 ANES, which reported a mean of 3.082 days (SD = 1.61).5. A potential confound of administering political knowledge questions on an Internet-based survey is that participants can search for the answers. The median score on the political knowledge questions was just over 50%, and this suggests that participants are not searching for the responses. However, if participants did look up the answers, this is not necessarily a problem as part of political knowledge is knowing where and how to find political information (Lupia and McCubbins Citation2000).6. The manipulation check was an open-ended question asking individuals what they thought was the purpose of the study. Ninety-three percent of those who answered this question correctly identified the study to be about politics or women. Excluding those who did not answer the question correctly does not change the substantive results and the seven percent who did not correctly identify the study purpose are included in these analyses.7. There are no differences in stereotype reliance between Democrats and Republicans in both the female and male candidate conditions.8. The categories for income are 1 = under $20,000, 2 = $20,000–34,999, 3 = $35,000–49,999, 4 = $50,000–74,999, 5 = $75,000–99,000, and 6 = $100,000. Because categories 5 and 6 have lower levels of participants I collapsed these two income groups.9. I refrain from including a "neither likely nor unlikely" option as the inclusion of this choice category may trigger social desirability pressures among participants who do not want to support a female candidate, but also do not want to be perceived as biased against women.10. Replicating these models with an ordinal logit estimation rather than an ordinary least squares regression shows the same set of findings. The stereotype and candidate gender interaction is significant, p < 0.10.