Feminist Identity,Attitudes Toward Feminist Prototypes,and Willingness to Intervene in Everyday Sexist Events

Why Is Feminist Identity Important?

Feminist identity is associated with important psychological and behavioral outcomes. People who identify as feminists display lower levels of heterosexism (Szymanski, 2004) and greater psychological well-being (Saunders & Kashubeck-West, 2006; though see Yoder, Tobias, & Snell, 2011, for contradictory findings), and lower body surveillance and shame (Hurt et al., 2007). Feminist identity is also associated with behaviors that can promote feminism, boosting self-efficacy to achieve goals (Eisele & Stake, 2008), and predicting increased willingness to engage in feminist activism (Nelson et al., 2008; Yoder et al., 2011), as well as willingness to engage in a range of feminist behaviors such as donating more to feminist charities and bringing up feminist topics in conversation (Redford, Howell, Meijs, & Ratliff, 2018). These findings show both protective and empowering factors associated with feminist identity.

Moreover, feminist identity cannot be replaced by, or reduced to, feminist beliefs (i.e., beliefs about gender equality). Although feminist identity has been conflated with feminist beliefs (Enns, 1997), equality beliefs alone do not necessarily result in willingness to call oneself a feminist (Fitz, Zucker, & Bay-Cheng, 2012; Williams & Wittig, 1997; Zucker & Bay-Cheng, 2010). Furthermore, consistent with social identity theory (Tajfel & Turner, 1979), identity confers benefits independent of beliefs; feminist identity is associated with outcomes such as charity donations and body satisfaction over and above, or better than, beliefs in accord with feminism (Murnen & Smolak, 2009; Redford et al., 2018; Yoder et al., 2011). Thus, the benefits of feminist identity are not derived from one’s beliefs; people must actually view themselves as feminists.

Feminist identity is associated with myriad positive outcomes, but to establish whether it actually contributes to efforts to achieve gender equality, feminist behaviors are of the greatest interest. Feminist behavioral outcomes have been studied in previous research—most often in formal, sporadic, and potentially burdensome forms (requiring money or time commitments)—such as participating in marches and donating to charities. However, this approach could overlook experiences of feminist behavior, which may be better reflected in reactions to everyday sexist events in interpersonal interactions. In addition to being understudied, such interactions are more frequent, accessible, and naturalistic than formal activism and feminist identity could be the key to such confrontation. One study of bystander reactions found feminist women were more likely to confront perpetrators of gender prejudice (Brinkman, Dean, Simpson, McGinley, & Rosén, 2015). In addition, perceptions of sexism as harmful promotes behavioral intentions to discuss and seek information about gender inequity (Cundiff, Zawadzki, Danube, & Shields, 2014), suggesting a potential role of perceived vulnerability to sexism in willingness to intervene. Thus, it could be that perceived vulnerability to sexism relates to willingness to act via feminist identity—the prediction we tested—or that vulnerability has a more direct link to willingness to intervene following instances of gender bias.

Feminist Identity and Feminist Prototypes

People who identify as feminists are willing to claim the label publicly (Aronson, 2003; Zucker, 2004). This personal identity becomes integrated with a social identity (Turner & Oakes, 1986), in which they see themselves as part of a group made up of other self-identified feminists (Burn, Aboud, & Moyles, 2000; Leaper & Arias, 2011). Because people strive for positive identities (Tajfel & Turner, 1979), seeing prototypical members of a group (e.g., feminists) as positive can relate to greater identification with that group. Thus, in the current research, we examined how much people identify as feminist in a model that incorporated their attitudes toward prototypical feminists, among other variables.

A prototype of a group is the most central, representative group member, the one who most easily and frequently comes to mind when thinking of the group (Rosch, 1973). Prototypes are rated on a variety of characteristics (Gerrard, Gibbons, Stock, Vande Lune, & Cleveland, 2005); for example, a person may think of the prototypical feminist as cool and fun or as uncool and boring. These traits can be positive or negative, resulting in a positive or negative attitude toward the prototype of a group. According to the prototype-willingness model of behavior (PWM; Gerrard, Gibbons, Houlihan, Stock, & Pomery, 2008), when opportunities arise to behave in group-relevant ways, people with more positive attitudes toward the group prototype are more willing to engage in those behaviors. For example, more positive attitudes toward “smoker” prototypes are associated with greater willingness to smoke (Gerrard et al., 2008), and positive prototypes of healthy eaters are associated with healthy eating (Gerrits, De Ridder, De Wit, & Kuijer, 2009). Prototypes of political groups work in the same way. For example, negative stereotypes of the “typical environmentalist” lead people to avoid affiliating with them and to have less motivation to adopt pro-environmental behaviors (Bashir, Lockwood, Chasteen, Nadolny, & Noyes, 2013).

Most prototype studies have only examined explicit attitudes, directly asking participants about how they see prototypes (e.g., “how smart is the prototypical healthy eater?”). Measures of explicit attitudes require that people are able and willing to report all aspects of their attitudes. However, people may be unable or unwilling to fully report their attitudes toward stigmatized or marginalized groups. For example, research has often shown that in studies involving race, self-reported explicit attitudes have significantly lower predictive validity than do implicit attitude measures (Greenwald, Poehlman, Uhlmann, & Banaji, 2009). Because feminists are often the target of stereotypes, backlash, and rejection (Anastosopoulos & Desmarais, 2015; Robnett, Anderson, & Hunter, 2012; Yeung, Kay, & Peach, 2013), explicit measures about feminists may not capture a full range of attitudes people hold.

Alternatively, measures of implicit attitudes can overcome some of the limitations of measures of explicit attitudes. They may access more gut-level responses, and either do not alert participants to what is being measured, or reduce participants’ control over their responses (De Houwer, 2006). Implicit attitudes have been found to have greater predictive power than explicit ones across a range of outcomes in socially sensitive domains including employment discrimination, administration of medical treatment, and nonverbal behavior in interpersonal interactions (Greenwald et al., 2009). Implicit attitudes (defined here as the outcomes of implicit measures) toward prototypes outperform explicit ones at predicting risky sun-exposure behavior (Ratliff & Howell, 2015), and implicit attitudes toward the prototypical environmentalist predicted environmentally-friendly behavior (Ratliff, Howell, & Redford, 2017). This power has only very recently been harnessed in research on prototypes of feminism, but findings are promising. One study found that implicit attitudes significantly related to feminist identification, controlling for both explicit feminist prototypes and egalitarian gender beliefs (Redford et al., 2018).

Implicit measures are especially appropriate for testing the PWM (Gerrard et al., 2008). The PWM focuses on willingness to engage in behaviors when opportunities arise rather than on a priori behavioral intentions. Such spontaneous, unplanned behavior (e.g., friendliness; Dovidio, Kawakami, & Gaertner, 2002) tends to be better predicted by implicit than explicit measures. Thus, including implicit measures not only enhances predictive power but is theoretically consistent with the types of behavior examined in PWM studies.

We were also interested in perceived vulnerability, another aspect of the PWM. Vulnerability in many health behavior models refers to a person’s perception of individual vulnerability to risks associated with a behavior (Gerrard et al., 2008) and, as such, is associated with behavioral willingness (Lewis, King, Litt, Swanson, & Lee, 2016). Although vulnerability typically reflects consequences associated with health behavior, we hypothesized that perceived vulnerability to sexism was important in our theorized model as well (see Figure 1). Similar constructs have been associated with feminism; for instance, in the previous research, women who perceived sexism to be pervasive were more likely to support confronting sexist behavior (Garcia, Schmitt, Branscombe, & Ellemers, 2009; Kahn, Barreto, Kaiser, & Rego, 2015). In addition, experiences of sexism have been associated with collective action (Nelson et al., 2008) and with feminist identity (Henderson-King & Stewart, 1997; Leaper & Arias, 2011; Zucker, 2004), although there is some debate about the direction of the relation between sexism and these other constructs. Moreover, general awareness of sexism is associated with feminist identity (Leaper & Arias, 2011). Given that all of these constructs reflect people’s awareness of sexism in some way, we expected that perceived vulnerability to sexism would be relevant both to feminist identity and to participants’ willingness to intervene.

OPEN IN VIEWER

Figure 1.

Model 1—hypothesized model.

We expected that self-reported perceived vulnerability would also reflect the interplay of identity and systemic oppression. Some people who reject the feminist label likely do so in part because of their lived experiences with such oppression. Historically, activists who occupy multiple marginalized positions have been alienated from the mainstream women’s movement, particularly in the United States (e.g., see Combahee River Collective, 1977/2015). As a result, some people may reject or reconfigure “White feminism” or feminism that they believe does not adequately address the intersectional, multifaceted, situated, and structural nature of identity. A recent study (Hoskin, Jenson, & Blair, 2017) found that people who defined feminism using intersectional principles were more likely to identify as feminists. And some individuals with egalitarian beliefs reject feminism for the opposite reason: Non-labelers who endorse neoliberal beliefs of individualism, competition, and meritocracy may find that feminism focuses too much on structural forces and collective identity (Fitz et al., 2012). Thus, while feminist identity appears deeply rooted in a complex social system, this complexity is not often explored in research on feminist identity.

The current research was designed to advance understanding of how implicit and explicit prototypes relate to feminist identity, in a model that also includes other important constructs associated with feminist identity, such as gender equality beliefs (i.e., feminist beliefs) and perceived vulnerability to sexism. We included gender equality beliefs in the model to assess the utility of prototypes over and above those beliefs, as demonstrated in previous studies (Murnen & Smolak, 2009; Redford et al., 2018; Yoder et al., 2011). We also included perceived vulnerability to sexism, which is associated with both PWM and feminist identity research, but which we conceptualized uniquely in this study (i.e., we created a new measure, described below). By including both gender equality beliefs and perceived vulnerability to sexism, we sought to examine their effects and also to test whether feminist identity is reducible to these other variables.

The Present Research

The purpose of the present research was to explore the role of implicit and explicit attitudes toward prototypes of feminists—the central, representative feminist that comes to mind when someone thinks of feminists as a group—in relation to feminist identity, and whether feminist identity in turn relates to people’s willingness to intervene when they encounter everyday sexist events. We examined these questions in a model that included other potentially important constructs: feminist beliefs and perceived vulnerability to sexism (see Figure 1). Because we conceptualized vulnerability uniquely in this study—that is, a new measure of feminist behaviors, as opposed to health behavior models which situate vulnerability as perceived conditional consequences of behavior (Gerrard et al., 2008)—we could not use the extant literature to situate the variable with total confidence. However, because we acknowledge the role of structural factors in feminist labeling, we decided to place vulnerability alongside other precursors of feminist identity (see Figure 1).

Participants and Procedure

This study was approved by our university’s institutional review board. We recruited participants through Prolific (http://www.prolific.ac), a crowdsourcing website focusing specifically on academic research. Prolific was ideal for our purposes for several reasons: The website prevents users from participating in the same study more than once; the site mandates a “minimum wage” consistent with our ethical principles; and the website collects broad demographic data about its participant pool, meaning that we could recruit by age, gender, and location without developing and administering a pre-screening survey. Because of its mandatory minimum payments, Prolific can be costlier than other tools (such as Amazon Mechanical Turk): at least £5/hour ($6.61/hour) for participants plus commission costs of 12.5% on each reward and 10 p ($0.13) per participant. Prolific is based at the University of Oxford, but because initial recruitment takes place primarily on social media or through university flyer campaigns, the participant pool boasts a high percentage of participants from the United States, as well as the United Kingdom, though anyone may join the website. Volunteers fill out extensive demographic questionnaires to qualify or disqualify them from available studies. Participant demographics can be explored through graphs on the website (https://www.prolific.ac/demographics); the pool is not particularly diverse in terms of race or ethnicity (with ∼70% identifying as Caucasian), but it does offer a wider age range than a typical university setting (18–80, with the average age in our sample at about 32 years old). A recent study (Peer, Brandimarte, Samat, & Acquisti, 2017) found that Prolific was more diverse than other crowdsourcing sites in terms of ethnicity and geographic location, although this was not the case in our study.

Peer and colleagues (2017) found that Prolific participants reported higher naiveté and lower degrees of dishonest behavior as compared to Amazon Mechanical Turk participants. In addition to Prolific’s protections against repeat submissions, we administered a two-step attention check adapted from Downs, Holbrook, Sheng, and Cranor’s (2010) study: Participants read a short email, ostensibly as a formality, followed by two questions of differing complexity about the information in the email, designed to ensure that participants were attending to the content of the survey.

Participants who met our eligibility criteria (women in the United States over the age of 18) chose to answer our survey after reading a short description: “In this research study, you will be asked to answer some questions about your attitudes, beliefs, and behaviors relating to gender and feminism. We will also ask you to do a sorting task on your computer.” They were compensated approximately $2.50 for their time. Data were collected from 440 adult U.S. women. We eliminated some participants for data trimming purposes (n = 4: one participant who identified as a trans man, one participant who was under 18, and two participants who failed the attention check) and removed incomplete submissions for a final sample of 428 (Mage = 32.39 years, standard deviation [SD] = 11.85). On a political orientation scale ranging from strongly conservative (1) to strongly liberal (7), participants reported somewhat more liberal than conservative beliefs (M = 5.4, SD = 1.55). For additional demographic characteristics, refer to Table 1.

OPEN IN VIEWER

Table 1.

Demographic Characteristics of Participants.

Characteristic	n	%
Race
White	337	78.7
East Asian	26	6.1
Black/African American	23	5.4
American Indian/Alaska Native	3	0.7
Native Hawaiian/Pacific Islander	1	0.2
Multiracial	22	5.2
Other/unknown	9	2.1
Ethnicity
Hispanic/Latina	18	4.2
Not Hispanic or Latino	402	93.9
Unknown	8	1.9
Sexual orientation
Heterosexual/straight	302	70.6
Bisexual	79	18.5
Lesbian/gay	16	3.7
Queer	15	3.5
Other	16	3.7
Highest level of education
Some high school	2	0.5
High school	34	7.9
Some college	115	26.9
Associate’s degree	38	8.9
Bachelor’s degree	155	36.2
Some graduate school	34	7.9
Master’s degree	37	8.6
JD	4	0.9
PhD	7	1.6
MBA	1	0.2
Unknown	1	0.2
Economic situation
Not enough money to get by	24	5.6
Barely enough money to get by	78	18.2
Enough money for needs, but no extras	161	37.6
More than enough money	146	34.1
Well-to-do	15	3.5
Extremely well-to-do	3	0.7
Unknown	1	0.2

Materials

Willingness to intervene in situations of everyday sexism

In order to assess willingness to engage in everyday feminist behaviors, we developed 6 items exemplifying hostile or benevolent sexist situations. We based our measure on benevolent and hostile sexism items that Glick and Fiske (1996) included in the Ambivalent Sexism Inventory (see Table 2 for item list). Participants rated their willingness to intervene in given situations on a response scale from 1 (very unwilling) to 7 (very willing). We averaged the items to create a single score of behavioral willingness. Internal consistency reliability for the current sample was α = .69.

OPEN IN VIEWER

Table 2.

Items in the Willingness to Intervene in Situations of Everyday Sexism Measure.

Item Text	Type of Sexism
1. Imagine that you’re at a party and someone tells a joke that is degrading to women. How willing would you be to speak out against this?	Hostile
2. Imagine that you are helping a friend move into her new apartment. You offer to help carry furniture but instead she said, “No, we’ll let the men do the heavy lifting; come and help me unpack the dishes.” How willing would you be to speak out against this?	Benevolent
3. Imagine that you learn that you make less money than a male co-worker for doing the same labor. How willing would you be to speak out against this?	Hostile
4. Imagine that you are at a job interview and your potential boss asks you if you plan to have children because he “doesn’t want to hire someone who will be quitting soon to be a stay-at-home mom.” How willing would you be to speak out against this?	Hostile
5. Imagine that you are at a restaurant and when you ask the waiter for the bill, he gives it directly to your male date instead of to you. How willing would you be to speak out against this?	Benevolent
6. Imagine that you are on a date with a male partner and he insists on opening the car and restaurant doors for you. How willing would you be to speak out against this?	Benevolent

Note. In future iterations of this measure, we plan to use the phrase “male friend” instead of “male date/partner” in Items 5 and 6.

Preliminary Analyses

All variables were moderately to highly correlated (correlations and descriptive statistics are presented in Table 3). Participants reported generally egalitarian attitudes about gender roles (M = 77.59, SD = 14.33) and moderately high identification with feminism (M = 5.03, SD = 1.76). We conducted all other analyses using path analysis in SPSS AMOS (Version 12). We also analyzed patterns of missing data for each variable. Less than 0.3% of all items in all cases were missing. About 91.6% of participants had no missing data whatsoever, and 63.6% of items had no missing data across all cases. No item had more than 1% missing values. Missing data were managed on the scale level with prorated mean imputation.

OPEN IN VIEWER

Table 3.

Zero-Order Correlations and Descriptive Statistics for Indicator Variables.

Variables	1	2	3	4	5	6
1. Implicit prototypes
2. Explicit prototypes	.32**
3. Gender equality attitudes	.25**	.35**
4. Feminist identification	.32**	.67**	.53**
5. Perceived vulnerability	.16**	.30**	.19**	.45**
6. Behavioral willingness	.14*	.37**	.36**	.44**	.34**
Mean	.25	4.87	77.59	5.03	4.37	4.49
Standard Deviation	.26	1.23	14.33	1.76	1.66	1.15

Note. *p ≤ .01. **p ≤ .001.

Testing Models

We first tested a model logically derived from our theory (Model 1, pictured in Figure 1), as discussed in the Introduction section. The χ² of this model was significant, χ²(4) = 33.58, p < .001, indicating poor fit. However, this statistic is sensitive to large sample sizes and assumes perfect multivariate normality; as a result, we examined additional fit statistics. We followed Byrne’s (2009) recommendations and looked to the comparative fit index (CFI; .96) and the root mean square error of approximation (RMSEA; .13, with a 90% confidence interval [CI] between .09 and .17, p of close fit [PCLOSE] < .001), as well as the non-normed fit index (NNFI; .83) and the standardized root mean square residual (SRMR; .05), for clarification; these results suggested mixed fit. Thus, we respecified the model according to the modification indices. We made two changes based on the modification indices and the standardized residual covariances, which were significant only for the relation between willingness to intervene and our measure of vulnerability (standardized residual covariance value of 2.81) and between willingness and gender role beliefs (standardized residual covariance value of 2.61). Drawing a direct path from gender role beliefs to behavioral willingness (M.I. = 8.90) and from perceived vulnerability to behavioral willingness (M.I. = 10.24) yielded Model 2, χ²(2) 5.83, p = .05, CFI = .99, NNFI = .96, SRMR = .02, RMSEA = .07, RMSEA 90% CI between <.001 and .13 with PCLOSE = .25; these changes improved the model, but the fit was still somewhat mixed. However, there were no useful modification indices.

We returned to the literature to assess the relations between the constructs. Given that our inclusion of perceived vulnerability is unique in this subject area, we focused on the possibility that it might need to be restructured. Although our original expectations constructed vulnerability as a precursor to feminist identity, we also acknowledged that vulnerability may directly relate to willingness rather than indirectly through feminist identity. The modification indices of Model 2 did support a direct link from vulnerability to willingness. Thus, we decided to reconceptualize vulnerability’s role somewhat: We eliminated the indirect path from vulnerability through feminist identity. This retained the direct paths from gender role beliefs and vulnerability to behavioral willingness in Model 2 but relocated the vulnerability construct such that a direct path ran from feminist identity to vulnerability (Model 3; see Figure 2). Model 3 fits the data well, χ²(5) = 7.93, p = .16, CFI = .99, NNFI = .99, SRMR = .02, RMSEA = .04. Although the RMSEA 90% CI is somewhat wide for this model (lower bound < .001 and upper bound = .08), it falls roughly within the values suggested by Browne and Cudeck (1993) for and in conjunction with the associated probability value (p = .62) and other fit indices; we believe that this fit is acceptable. In terms of local fit, none of the standardized residual covariances were large enough to cause concern: Values are listed in Table 4. Standardized total and indirect effects are listed in Tables 5 and 6, respectively. Because of the small effects of the implicit measure relative to other variables, we tested a version of this well-fitting model without the IAT included; the resulting model, Model 4: χ²(3) = 6.79, p = .08, CFI = .99, NNFI = .98, SRMR = .03, RMSEA = .05, fits no better. As a result, we chose to retain implicit attitudes, as in Model 3.

OPEN IN VIEWER

Figure 2.

Standardized coefficients for Model 3—adjusted final model, χ²(5) = 7.93, p = .16, CFI = .99, NNFI = .99, SRMR = .02, RMSEA = .04. e1, e2, and e3 represent error terms. *p < .05. ***p ≤ .001.

OPEN IN VIEWER

Table 4.

Standardized Residual Covariances for Model 3.

Variables	1	2	3	4	5	6
1. Implicit prototypes	<.001
2. Explicit prototypes	<.001	<.001
3. Gender equality attitudes	<.001	<.001	<.001
4. Feminist identification	<.001	<.001	<.001	<.001
5. Perceived vulnerability	.31	−.15	−.98	<.001	<.001
6. Behavioral willingness	−.41	1.34	−.18	<.001	−.18	−.05

OPEN IN VIEWER

Table 5.

Standardized Total Effects for Model 3.

Variables	1	2	3	4	5	6
1. Implicit prototypes
2. Explicit prototypes
3. Gender equality attitudes
4. Feminist identification	.07	.54	.32
5. Perceived vulnerability	.03	.24	.14	.45
6. Behavioral willingness	.02	.18	.30	.34	.19

OPEN IN VIEWER

Table 6.

Standardized Indirect Effects for Model 3.

Variables	1	2	3	4	5	6
1. Implicit prototypes
2. Explicit prototypes
3. Gender equality attitudes
4. Feminist identification
5. Perceived vulnerability	.03	.24	.14
6. Behavioral willingness	.02	.18	.11	.08

Limitations

The primary limitation in the current research is demographic: Our sample consisted primarily of heterosexual White women whose average political orientation was fairly liberal. Because women with diverse racial, ethnic, and class backgrounds often have different feminist values, definitions, and identities (Harnois, 2005; Moradi, Subich, & Phillips, 2002; Robnett & Anderson, 2017), a complete model of feminist identity and behaviors would likely account for these factors. Unfortunately, our sample did not allow for statistical comparisons by race or ethnicity, and tests of mean differences comparing economic groups yielded no significant results; such nuance might be more evident in a more diverse sample. We specifically targeted women for this study because two of our variables of interest (behavioral willingness and vulnerability to sexism) would need significant restructuring to be applicable to men; however, in future studies, researchers may try to validate our results among men and people with non-binary gender identities.

In addition, not all constructs contributed substantially to the model, the contribution of implicit attitudes was relatively small compared to those of explicit attitudes and gender equality attitudes. It is possible that, because our participants primarily self-identified as liberal, generally positive opinions of feminists minimized the potential disparity between explicit and implicit attitudes, thus reducing the usefulness of the IAT. Our implicit measure of feminist prototypes has only been used in one other study (Redford et al., 2018), and we hope that in the future researchers will refine and validate it. However, the model without implicit attitudes fits no better than the model that included implicit attitudes. Therefore, this study highlights a novel piece that is missing in other conceptualizations of feminist identity; although it is a small contribution, it is potentially an integral one. In short, we can better understand feminist identity—and relevant behavior—by understanding how women feel about feminists.

There are other limitations regarding our model. When the hypothesized model yielded a mixed fit, we needed to identify a model that fit the data better; our model configuration is novel and requires replication. Although we acknowledge that the changes made in Model 3 are consistent with existing literature, the fact that the initial model required respecification necessitates caution. Because modifications in Model 3 were made in response to analyses with the current data set, it is possible that the results may represent spurious patterns particular to our sample (Bollen & Noble, 2011). There are undoubtedly also variables missing from the model. For example, Kaysen and Stake (2001) predicted activism in part through agentic self-esteem, although it is possible that such a relation would be complicated by our use of everyday feminist behaviors instead of formal activism. We also acknowledge that the two measures designed for this study (perceived vulnerability to sexism and willingness to intervene in everyday sexist situations) are limited by a lack of replication and validation. Reliability for the vulnerability measure was good (α = .81), but not excellent, and may yield biased results because of its direct use of the term “sexism.” Reliability for the measure of willingness to intervene was low but acceptable (α = .69). Reliability was not improved by separating hostile sexism and benevolent sexism subscales, nor did an item-by-item analysis reveal any one item explaining the level of internal consistency. We chose these 6 items to reflect a range of sexist events that might occur to a participant in everyday life, but the items may not be measuring the same construct. Finally, because our study involved self-report, we are limited in our ability to extend our results to behavioral outcomes.

Future Directions for Research

Researchers may consider replicating our findings in a more diverse sample; a replication may also resolve concerns about model respecification. We also hope that other researchers will consider incorporating the variables we included into their own feminist identity models: The willingness to intervene construct, though in need of psychometric refinement, reflects the everyday sexist interactions that feminists regularly face (rather than focusing on activist work, as in past research), the vulnerability construct includes assessment of women’s personal and individual relation with systemic oppression, and the feminist prototypes measures can add nuance to investigations of feminist identity by considering an additional layer of attitudinal motivation. The measure we used in the present research to assess willingness to intervene in a sexist situation could be improved by qualitative research that expands and clarifies the specific sexist scenarios that elicit intervention behaviors. We also think that there is potential to investigate differences in feminist identity and behavior in privileged or disadvantaged groups. For example, does privilege make White, cisgender, heterosexual, middle- or upper-class feminists perceive less personal risk? Does that perception inspire fewer feminist behaviors? Alternatively, does it reduce the risk associated with confrontations of sexism, thereby increasing feminist behaviors? Such privilege might allow some feminists to survive, or ignore the risk of, job loss resulting from interpersonal confrontations at work. Is this differentially effective for everyday interactions compared to activist organizing? In a diverse sample, we believe our model could help characterize the increasing visibility of risk of marginalized populations.

Practice Implications

By intervening when faced with sexism, feminists may help achieve greater gender equity. Such interventions may reduce subsequent bias (for a similar study related to racism, see Czopp, Monteith, & Mark, 2006) and increase men’s attempts to compensate for their sexism in later interactions (Mallett & Wagner, 2011). However, the social costs of intervening, such as being evaluated negatively or punished, often prevent women from doing so (Nicole & Stewart, 2004; Swim, Eyssell, Murdoch, & Ferguson, 2010; Swim & Hyers, 1999). Despite the possible benefits, researchers rarely study confrontations of sexist behaviors (Becker, Zawadzki, & Shields, 2014).

Our results suggest that consciousness-raising may be effective in promoting feminist action. For example, educators or interventionists hoping to promote change may want to promote a more favorable “prototypical feminist” (thereby increasing feminist identification) and also may provide information about vulnerability to sexism (in order to motivate women to engage in feminist behaviors). Women’s Studies courses could be an ideal venue for interventions involving prototypes and targeted consciousness-raising, as students enrolled in those courses display increases in both feminist consciousness and feminist identification (Henderson-King & Stewart, 1999), and Women’s Studies students who are more aware of sexism demonstrate greater activist intentions (Stake, 2007). Our research speaks to one pathway through which this may occur.

Conclusions

The current research illustrates how favorable attitudes toward prototypical feminists relate to feminist identity, which in turn relates to willingness to engage in feminist behaviors, especially for women who feel that they are personally vulnerable to sexism. Our findings show the importance of awareness of sexism, attitudes toward feminists, and feminist identity for reducing sexist interactions. To live in a world with fewer gendered constraints on their ambitions, outcomes, and opportunities, yes, women do need feminism.