Introduction

From prior research we know that ambivalent sexism contributes to legitimizing male dominance and women’s lower status in society (Connor et al., 2017). Gender-based inequality continues to be widespread in many areas (e.g., economic empowerment, violence, sharing of housework and childcare) (UN Women, 2021) and is well reflected, for instance, in women’s lower presence in the labor market. Even when women do enter the job market, they face greater difficulties in accessing quality employment opportunities (ILO, 2022). Ambivalent sexism combines negative (i.e., women are perceived as inferior and unqualified) and positive (i.e., virtuous women deserve to be placed on a pedestal) attitudes. This mixture of hostility and subjective benevolence stems from the conjunction of female-male interdependence, patriarchy, and gender differentiation through roles and stereotypes (Glick & Fiske, 1996, 2001). This is relevant because people still accept and share unjustifiable beliefs about the traits considered to be typical of or ideal for each sex in society (e.g., men are attributed competence, whereas women are depicted as being nice but less competent than men) (Glick & Fiske, 1996). In addition, these beliefs are incorporated in individuals’ self-descriptions and guide their attitudes and behavior (Wood & Eagly, 2009).

Yet a complication arises because it is difficult to detect ambivalent sexism and, consequently, to establish how these gender-typed personal traits are connected to sexist attitudes. This may be because the main method used to assess people’s ambivalent sexism is the Ambivalent Sexism Inventory (ASI) (Glick & Fiske, 1996). As the ASI is a self-report instrument, the likelihood of participants feeling pressured to respond in a socially desirable way may be high (Bragg, 2011), thus increasing the risk of response bias. However, as far as we know, no studies explicitly examine the influence of social desirability in the association between self-ascribed gender traits and ambivalent sexism. This complication is concerning because people display attitudes and behave in ways that are consistent with their gender identity based on gender-stereotypic traits (Wood & Eagly, 2009) that could be harmful to women. Moreover, because the hostile dimension of ambivalent sexism is socially censured, individuals tend to hide it, which makes it difficult to identify. Benevolent sexism is regarded as relatively inoffensive or even romantic, and thus reinforces the diffusion and acceptance of sexist ideology (Connor et al., 2017).

We address this concern by attempting to fill this gap in the literature, specifically by assessing the effect of social desirability (i.e., the degree to which people seek social approval through culturally acceptable responses) (Crowne et al., 1960) on the association between self-described gender traits (i.e., the individual’s self-ascription of culturally defined masculine and feminine personality traits) (Bem, 1974), and ambivalent sexist attitudes (i.e., a manifest antipathy toward women together with an apparently positive but patronizing attitude toward them) (Glick & Fiske, 1996). Moreover, the research has usually studied the linkage between each gender trait dimension and ambivalent sexism separately (e.g., Lameiras et al., 2007; Spence & Buckner, 2000), rather than exploring a possible moderation of one dimension on the other. For this reason, we examine whether the modulating effect of social desirability in the linkage between one gender trait dimension and sexist attitudes is, in turn, moderated by the respondents’ scores in the other dimension. This paper contributes to the literature by expanding the understanding of what causes the widespread diffusion and acceptance of ambivalent sexist attitudes, which are not explicit in social relations, but are evidenced in women’s lower social status compared to men.

Theoretical framework

Ambivalent sexism

Prejudice against women is not so much an intense antipathy toward them as a well-defined ambivalence (Glick & Fiske, 1996). Consistent with ambivalent sexism theory, the construct of sexism (Glick & Fiske, 1996, 2001) has two dimensions (i.e., hostility and benevolence), each with its own connotations. Hostile attitudes denote clear resentment of women, and one way of justifying men’s dominant position in society is by treating women in an offensive and hostile manner. Women are perceived as trying to usurp male power through their sexuality, by claiming discriminatory treatment, or through feminist actions (Connor et al., 2017).

By contrast, benevolent sexism expresses apparently positive attitudes toward women as a way of justifying male domination. It stresses the male role of protector and provider, which, through chivalry, puts women on a pedestal (Mastari et al., 2019). It is a form of paternalism aimed at protecting women (i.e., chivalrous behavior toward women), complementary gender roles (i.e., stereotypical roles for women and men), and romantic heterosexual intimacy (i.e., the assumption that a man is incomplete without a woman and vice versa). It idealizes heterosexual relationships and considers women to be the perfect complement to men; however, it also assumes that women are weak and need a man to protect them, thereby reinforcing their lower status in society (Connor et al., 2017). Although this attitude may seem favorable, such benevolence is in fact prejudicial since it strengthens traditional stereotypes and enhances male dominance (e.g., men are providers and women depend on them). This frequently has damaging consequences for women, such as slowing down their progress in the workplace and undermining their career ambitions and their collective resistance to gender inequality (Glick & Raberg, 2018). Benevolent sexists idealize women who adopt a traditional role rather than denigrating those who follow a nontraditional path (Glick & Fiske, 2011). This attitude seems to be more prevalent and socially accepted than hostile sexism, even among women (Hideg & Shen, 2019; Mastari et al., 2019).

Ambivalent sexism does not induce any cognitive conflict because women are divided into polarized subgroups: women who adopt nontraditional roles (e.g., career women) deserve hostile treatment, while others (e.g., homemakers) merit pedestal status. Hostile and benevolent sexist attitudes are positively associated (Glick & Fiske, 1996), and so concur to validate and preserve the patriarchy and traditional gender roles.

Stereotypical gender traits

The expression of sexist attitudes may derive from the assimilation of traditional gender roles, that is, the beliefs shared by members of a given society about what traits are appropriate for women and men. Gender roles motivate people to adapt to these shared beliefs by accepting normative expectations and incorporating them as personal standards that guide their own behavior (Wood & Eagly, 2009). A pivotal approach to conceptualizing gender stereotypes as well as gender self-concepts distinguishes between masculine and feminine traits. Traits labeled as instrumental (e.g., activeness, assertiveness, or independence) are agentic in nature and are stereotypically viewed as masculine. In turn, the so-called expressive traits (e.g., kindness, gentleness, or tendency to be emotional), which are communal attributes, are considered feminine (e.g., Abele, 2003; Bem, 1974). Agency involves achievement orientation (e.g., competence), disposition to take charge (dominance), autonomy (e.g., independence), and rationality (e.g., analytical skills). Communality implies interest and concern for others (e.g., kindness), a tendency to affiliate with others (e.g., warmth), deference (e.g., obedience), and emotional sensitivity (e.g., intuition) (Heilman, 2012).

Bem’s (1974) approach is central to the research on gender identity grounded in gender-stereotypical personality traits. She developed the Bem Sex-Role Inventory (BSRI), which encompasses personality traits considered as stereotypically masculine (e.g., dominant, hard-hearted, aggressive) or stereotypically feminine (e.g., understanding, sensitive to others’ needs, compassionate) (Bem, 1974). The inventory has been widely used in different cultural settings to measure gender role stereotyping. Men usually self-attribute agentic traits, while women define themselves in more communal terms (Wood & Eagly, 2009), although these patterns have shifted over time. Communal attributes, such as helpfulness, warmth, kindness and understanding, are generally regarded as more desirable and are therefore assessed more positively than agentic characteristics, such as independence and self-confidence. Consequently, women could be viewed more favorably than men. Some researchers have also documented the “women-are-wonderful” effect, in which more positive and valued qualities are associated with women than with men (Langford & MacKinnon, 2000).

Twenge’s (1997) meta-analysis showed how levels of communion have remained higher among women than men. In turn, agency has steadily increased among both men and women (one of the measures used was the BSRI), and the gap between the two sexes is shrinking (see also Spence & Buckner, 2000). Some studies using samples of university students have found no differences between men’s and women’s agentic traits (Abele, 2000). Moreover, social changes, such as more female participation in the public domain and women’s incorporation in the labor market, have modified gender stereotypes of women in Latin America and the U.S.: women showed higher levels of some masculine traits (e.g., competitive, physically energetic) and lower levels of some feminine traits (e.g., nurturing, attractiveness) (Diekman et al., 2005). However, a recent meta-analysis of changes in gender stereotypes since the mid-twentieth century evidenced a clear growth in the attribution of communal traits to women in relation to men while their agentic traits remained stable. In the same vein, men retain their agency advantage (Eagly et al., 2019).

Stereotypical gender traits, ambivalent sexism, and social desirability

Empirical evidence reveals sex differences in ambivalent sexist attitudes. Men tend to obtain higher scores than women on the benevolent and hostile sexism subscales, but the differences are more extreme for the hostile than for the benevolent dimension, with men consistently scoring higher on the hostile attitude (Becker & Wright, 2011; Chen et al., 2009; Glick & Fiske, 1996). Women are less likely to reject attitudes and beliefs that could be beneficial to them, such as the protection associated with benevolent sexism (particularly in the most sexist cultures) (Glick & Fiske, 2001), than overtly sexist attitudes that are hostile toward them (Glick & Fiske, 1996). Research has also found that masculine identification (i.e., collective identity as a man) correlates positively with hostile rather than benevolent sexism (Glick et al., 2015).

In their study using specific scales that measure self-ascribed expressive and instrumental traits, Lameiras et al. (2007) obtained negative correlations between the expressive and hostile sexism scales among women, as measured by the Personal Attributes Questionnaire. In the case of men, expressiveness correlated positively with benevolent sexism. Using the BSRI, Spence and Buckner (2000) also found that women’s scores on the expressive scale correlated negatively with hostile sexist attitudes. This outcome might be due to an expressive dimension per se: women with a focus on others could be predisposed to reject hostility directed at themselves or other women. These authors also found significantly lower hostile sexist attitudes among men who self-reported as having high expressive traits. Possibly because they are counter-stereotypic, men who self-attributed more expressive traits were likely to respond favorably to women and be more sensitive to their cause. Therefore, previous empirical evidence suggests, in general terms, that men are likely to have higher scores than women in ambivalent sexism, particularly in the hostile dimension. In addition, masculinity and instrumental traits are positively linked to hostile sexism. In turn, expressive traits are related positively to benevolent attitudes, but negatively to hostile sexism. However, the literature provides neither a clear picture of the findings, nor a robust linkage between self-ascribed gender attributes and sexism (Spence & Buckner, 2000).

This lack of clarity may be due to social desirability, which can be understood as the tendency of individuals to bias their responses so they appear favorably to others and reflect what they consider will win others’ approval, or avoid their disapproval (Crowne & Marlowe, 1960). These desirable responses may appear in socially sensitive issues such as sexist attitudes. In fact, the likelihood that participants will feel under pressure to answer in a socially approved way may be high in studies about ambivalent sexism (Bragg, 2011). Individuals may want to mask their sexist attitudes, or they may simply be unaware of their actual prejudices (Lorenzi-Cioldi & Kulich, 2015). Social desirability motivates us to express ourselves publicly according to the expectations that the social group has about us; hence, displaying hostile sexism would not be socially desirable. Benevolent sexism is more socially accepted than hostile sexism (Hideg & Shen, 2019; Mastari et al., 2019), so it is less likely to be regarded as prejudicial (i.e., holding benevolent attitudes might not be perceived as dangerous or harmful); consequently, desirability might be higher in this case. Chisango et al.’s (2015) research on this topic found that women were more likely to encounter hostility from their male partners in private situations, whereas they experienced more benevolent sexism in public settings, thus confirming social disapproval of hostile sexism, but social acceptance of benevolent sexism.

Ambivalent sexism entails a combination of antipathy toward women who are seen as a potential threat to men’s power and, at the same time, affection for women who embrace traditionally feminine roles (Glick & Fiske, 1996; Mastari et al., 2019). Self-attributed stereotypical gender traits seem to be involved in shaping such sexist attitudes: stereotypically feminine characteristics appear to be more connected to benevolent sexism (Lameiras et al., 2007), whereas masculine attributes are associated more with hostility toward women (Glick et al., 2015). However, the empirical evidence provides no well-defined picture of the association between individuals’ self-perceptions of gender attributes and their sexist attitudes (Spence & Buckner, 2000). We argue that this may be because social desirability could motivate participants to hide their negative attitudes (impression management), or conversely, they may believe that holding certain protective attitudes toward women is positive. As mentioned above, benevolent sexist attitudes meet with greater social acceptance than hostile attitudes (Chisango et al., 2015; Hideg & Shen, 2019; Mastari et al., 2019). As far as we know, little research has explored the influence of participants’ social desirability in the association between self-ascribed gender attributes and sexism. In this study we try to fill this gap in the research. In addition, when previous studies have analyzed the association between one gender trait dimension and ambivalent sexism, the possible influence of the other gender trait dimension in that relationship is not included, which could also bias the findings. Here we examine the possible influence of both dimensions.

Our purpose in the present research is to test a moderated moderation, specifically the moderating role of social desirability in the association between one gender trait dimension and ambivalent sexism where the other gender trait dimension is also a moderator. We argue that participants who self-describe as high stereotypically feminine adopt higher benevolent attitudes toward women (e.g., perception of needing male protection and meriting “pedestal status”) (Lameiras et al., 2007), particularly when those individuals assume low stereotypically masculine attributes (e.g., masculine traits such as hard-hearted would not seem compatible with benevolence), and tend to offer high socially desirable responses (i.e., benevolent sexism is more socially accepted as it appears to be potentially rewarding) (Becker & Wright, 2011; Chisango et al., 2015). We expect more benevolent sexism among these participants than among the individuals who share the same gender traits profile (i.e., high feminine and low masculine traits), but do not respond in a socially desirable way. Moreover, we argue that the individuals who self-ascribe more stereotypically masculine attributes also express more hostility toward women (e.g., women are believed to use various channels to usurp male power, such as their sexuality or claiming discriminatory treatment) (Glick et al., 2015), particularly when those participants score lower in stereotypically feminine attributes (Spence & Buckner, 2000), and they are indifferent to the way others perceive them. We expect higher hostile sexism among these participants than among those who also self-ascribe high masculine and low feminine traits, but prefer to offer socially acceptable responses. Because hostile sexism now meets with social disapproval (Chisango et al., 2015; Hideg & Shen, 2019; Mastari et al., 2019), people may want to mask their hostility (Lorenzi-Cioldi & Kulich, 2015). Hence, when an individual dares to explicitly demonstrate their hostility toward women, it is likely due to their low tendency to offer socially desirable responses. Figures 1 and 2 depict these two conceptual models.

Fig. 1
figure 1

Benevolent sexism conceptual model

Full size image
Fig. 2
figure 2

Hostile sexism conceptual model

Full size image

Accordingly, we expect two three-way interaction effects in the relationships between self-ascribed masculine and feminine traits, social desirability, and ambivalent sexism, which leads us to formulate two hypotheses:

  • Hypothesis 1: Masculine traits and social desirability in the feminine traits–benevolent sexism relationship: high self-ascribed feminine participants score higher in benevolent sexism when they display low masculine traits and high socially desirable responses, as compared to those individuals with the same gender traits profile, but low scores in social desirability.

  • Hypothesis 2: Feminine traits and social desirability in the masculine traits–hostile sexism relationship: high self-ascribed masculine participants score higher in hostile sexism when they show low feminine traits and low socially desirable responses, as compared to those individuals with the same gender traits profile, but high scores in social desirability.

Method

Participants

Prior to selecting the sample, we defined the statistical values to establish the power of analysis. Based on α = 0.01, a theoretical statistical power of 0.95 and an effect size of 0.15, the estimated sample size needed for the study was 189 participants for a linear multiple regression with a fixed model of six predictors. Our study exceeded this number of participants: 294 bachelor’s degree students from a university in eastern Spain took part in this research, conducted in 2021. Their sociodemographic profile reflected the middle socioeconomic status typical of Spanish public universities. Their informed consent was obtained, and they completed a questionnaire that included a set of demographic questions on sex, date of birth, degree studied, and employment situation. The students were aged between 21 and 51 years (M = 25.32, SD = 3.75). Roughly the same number of women and men took part: n = 148, 50.3%, women; n = 146, 49.7%, men. Participants were students of legal and social sciences such as management (n = 160, 54.4%), health sciences such as psychology (n = 113, 38.4%), engineering and architecture (n = 7, 2.4%), science (n = 4, 1.4%) and arts and humanities (n = 10, 3.4%); 22.1% (n = 65) were working at the time the study was conducted.

Procedure

The students responded to an e-mail inviting them to participate in a study on leadership and organizations that did not involve medical experimentation; approval for the research was granted by the university research ethics committee (CD/50/2021). The e-mail also provided information on the participation conditions (timetable, location, and guarantee of anonymity in the data treatment). Those who agreed to participate provided their informed consent before completing the questionnaire individually in a 10-minute session. The confidentiality and anonymity of responses were fully guaranteed.

Measures

Self-ascribed gender traits

We assessed self-ascribed gender traits using the reduced version of the Bem Sex-Role Inventory (BSRI; Bem, 1974) adapted to the Spanish context (López-Sáez & Morales, 1995). According to Cuadrado (2004), this recent version is a useful adaptation of the instrument to Spanish society in that it includes new negative traits traditionally attributed to women and men in Spanish culture. It consists of 18 adjectives (nine are stereotypically masculine and nine, feminine). Participants evaluated the extent to which they associated each adjective with themselves on a 7-point Likert scale ranging from 1 (never) to 7 (always). Following previous research, we used the mean scores of the masculine traits scale and feminine traits scale to obtain two separate gender role scores: a masculine score and a feminine score. The BSRI presented adequate psychometric properties: Cronbach’s alpha coefficient for the feminine traits scale was 0.77 and for the masculine traits scale, 0.74.

Ambivalent sexism

We used an adaptation to the Spanish context (Expósito et al., 1998) of the Ambivalent Sexism Inventory (ASI) (Glick & Fiske, 1996) to assess sexist beliefs and attitudes (Glick & Fiske, 2011). The scale contains 22 items, each of which was evaluated on a scale ranging from 1 (totally disagree) to 6 (totally agree). All item responses were scored in the same direction: higher scores on the scale denote higher levels of sexist judgments.

The instrument is composed of two subscales, each with 11 items: the benevolent and the hostile sexism subscales. Benevolent sexism (BS) refers to beliefs that are seemingly positive and predisposed to intimacy-seeking or prosocial behaviors (e.g., “Women should be cherished and protected by men”). By contrast, hostile sexism (HS) involves unfavorable beliefs and antipathy toward women (e.g., “Women try to gain power by controlling men”). Cronbach’s alpha coefficients were 0.88 for HS and 0.82 for BS.

Social desirability

The Marlowe-Crowne Social Desirability Scale (M-C SDS) Short Form adapted to Spanish populations by Ferrando and Chico (2000) was used to assess the participants’ desire to present themselves favorably (Reynolds, 1982). The short form of this scale has 13 items (e.g., “When I make a mistake, I am always willing to admit it”), which are rated as true or false. Individuals’ responses are labeled as socially desirable (0) or not socially desirable (1) and are added up to give a global score ranging from 0 (all socially desirable answers) to 13 (no socially desirable answers): the higher the score, the lower the level of social desirability. The validity of this scale has been widely supported in previous work (Reynolds, 1982). We obtained a Cronbach’s alpha of 0.61 in this study.

Control variables

The participants reported sex (operationalized as a dummy variable with two values: -0.5 = male and 0.5 = female), age (in years) and employment situation (measured as a dummy variable with two values: -0.5 = working at the time of the study and 0.5 = not working). We incorporated these factors, which have often been used as control variables in previous research since it has been found, for instance, that ambivalent sexism differs across age (e.g., Hammond et al., 2018) or that women describe themselves as relatively expressive/communal, while men generally describe themselves more in instrumental/agentic terms (e.g., Bem, 1974; Wood & Eagly, 2009).

Statistical analyses

To test our hypotheses, we used Model 3 (moderated moderation) of the SPSS PROCESS macro (Hayes, 2018). Model 3 can be used to estimate the effect of several moderators on a dependent variable (DV), since the results obtained will indicate whether the moderation of X (predictor variable) on Y (criterion or dependent variable) by M (moderator variable) depends on W (moderator variable). Despite instances of the predictors not exhibiting significant associations with the DV in the previous regressions, we investigated moderation following Hayes’s recommendations (2018). Two sets of analyses were run: one analysis with the benevolent sexist dimension as the criterion, and another with hostile sexism as the criterion or dependent variable. Analyses were mean centered.

Results

Preliminary analyses

We conducted descriptive and correlational analysis using SPSS software release 26 (IBM Corp.) as our preliminary analysis. Table 1 reports the results, which showed negative and significant relationships between participants’ self-ascribed feminine and masculine traits and hostile sexism, but positive ones with social desirability. The higher the undergraduates’ scores in feminine traits, the lower their masculine traits and their hostile sexism, but the higher their social desirability.

Table 1 Means, Standard Deviations, Cronbach’s Alpha, and Zero Order Pearson’s Correlations among the Variables
Full size table

Undergraduates’ self-ascribed masculine traits linked positively to both hostile and benevolent sexism, but their association with social desirability was negative; that is, the higher the undergraduates’ scores in masculine traits, the higher both their hostile and benevolent sexist attitudes, but the lower their social desirability.

Results also revealed a positive association between hostile sexism and benevolent sexism, but a negative link between hostile sexism and social desirability. Finally, it is interesting to note that participants’ sex was significantly related to all the variables included in this study, except age. Women showed higher levels of self-ascribed feminine traits and greater social desirability than men. Men’s scores were higher than women’s in both benevolent and hostile sexist attitudes and in self-ascribed masculine traits.

Predicting benevolent sexism

To test Hypothesis 1, we performed a moderated moderation analysis using Model 3 of Hayes’s (2018) SPSS PROCESS macro, taking benevolent sexism as the criterion, self-ascribed feminine traits as the predictor variable, and social desirability and masculine traits as moderators. The main potential influences of the control variables––participants’ sex, age, and employment situation––were entered as covariates. Table 2 presents the findings for benevolent sexism.

Table 2 Moderated Moderation on Benevolent Sexism Results (Model 1)
Full size table

The model was statistically significant (R2 = 11%; F = 3.327, p < 0.001). The control variables sex and age had a significant influence on benevolent sexism (β = -0.26, SE = 0.11, t = -2.35, 95% CI[-0.48, -0.04], p = 0.019, and β = -0.04, SE = 0.01, t = -3.20, 95% CI [-0.06, -0.01], p = 0.001), respectively: women and older undergraduates exhibited lower levels of benevolent sexism. We obtained a three-way interaction among feminine traits, social desirability and masculine traits, with a near-significant trend (β = 0.58, SE = 0.30, t = 1.92, 95% CI [-0.01, 1.1], p = 0.056), accounting for 1.0% of the variance in benevolent sexism (F(1, 293)=3.68, p = 0.056). Therefore, Hypothesis 1 was not supported. However, the conditional effects of femininity on benevolent sexism at different levels of masculinity and social desirability displayed in Table 3 showed a significant effect for those individuals low in both masculinity traits and in social desirability.

Table 3 Conditional effects of feminine traits–benevolent sexism at different levels of social desirability and masculinity traits
Full size table

Figure 3 plots the linkage between feminine traits and benevolent sexism at low and high levels of social desirability and masculine traits (i.e., 1 SD below and above the mean). Social desirability and masculine traits were conditionally related in the association between feminine traits and benevolent sexism. At low levels of social desirability, benevolent sexism increased as feminine traits rose for the individuals with low masculine traits (b = 0.25, p = 0.04), whereas the relationship between the two variables was insignificant for those with high masculine traits (b = -0.15, p = 0.19). In contrast, when social desirability was high, the association between feminine traits and benevolent sexism was insignificant at both low and high levels of masculine traits (b = 0.21, p = 0.059; b = 0.18, p = 0.16).

Fig. 3
figure 3

Moderation effects of social desirability and masculine traits on the relationship between feminine traits and benevolent sexism

Full size image

Predicting hostile sexism

Hypothesis 2 was also tested using Model 3 (moderated moderation) of the SPSS PROCESS macro by Hayes (2018), taking hostile sexism as the criterion, masculine traits as the predictor variable, and social desirability and feminine traits as moderators. The main potential influences of the control variables, namely participants’ sex, age, and employment situation, were entered as covariates. Table 4 reports the findings for hostile sexism.

Table 4 Moderated Moderation on Hostile Sexism Results (Model 2)
Full size table

This model was statistically significant (R2 = 21%; F = 7.66, p < 0.001). The control variables sex and age had a positive and significant influence on hostile sexism (β = -0.51, SE = 0.10, t = -4.73, 95% CI [-0.72, -0.29], p < 0.001, and β = -0.03, SE = 0.01, t = -2.31, 95% CI [-0.05, -0.004], p = 0.02), respectively: women and older undergraduates exhibited lower levels of hostile sexism. Participants’ masculine traits significantly predicted their hostile sexist attitudes (β = 0.12, SE = 0.06, t = 1.97, 95% CI [0.005, 0.93], p = 0.05). Social desirability had a direct and negative effect on hostile sexism (β = -0.68, SE = 0.24, t = -2.74, 95% CI [-1.16, -0.19], p = 0.006) in that the undergraduates with lower social desirability exhibited higher hostile sexism. We did not obtain a significant three-way interaction effect between masculinity, social desirability, and femininity, so Hypothesis 2 was not verified, although there were two significant two-way interactions in the same expected direction.

We found that social desirability moderated the effect of masculine traits on hostile sexism (β = -0.56, SE = 0.27, t = -2.07, 95% CI [-1.09, -0.02], p = 0.039). Figure 4 plots the relationship between masculine traits and hostile sexism at low and high levels of social desirability (i.e., 1 SD below and above the mean). At low levels of social desirability, hostile sexist attitudes grew as masculine traits increased (b = 0.25, p = 0.002). In contrast, when social desirability was high, the association between masculine traits and hostile sexism was insignificant (b = 0.001, p = 0.98).

Fig. 4
figure 4

Relationship between masculine traits and hostile sexism at low and high levels of social desirability

Full size image

We also found a statistically significant two-way interaction between masculine and feminine traits in the prediction of hostile sexist attitude (β = -0.17, SE = 0.06, t = -2.66, 95% CI [-0.30, -0.04], p = 0.008). Figure 5 plots the relationship between masculine traits and hostile sexism at low and high levels of feminine traits (i.e., 1 SD below and above the mean). Also, in this case we see that at low levels of feminine traits, hostile sexist attitudes intensified as masculine traits increased (b = 0.29, p = 0.001). In contrast, when the feminine traits score was high, the association between masculine traits and hostile sexism was insignificant (b = -0.02, p = 0.79).

Fig. 5
figure 5

Relationship between masculine traits and hostile sexism at low and high levels of feminine traits

Full size image

Discussion

We examined whether masculine traits moderated the modulating influence of social desirability in the association between feminine traits and benevolent sexist attitudes, and whether feminine traits moderated the modulating effect of social desirability in the linkage between masculine traits and hostility. The first point of interest is that levels of benevolent and hostile attitudes are lower among women and older participants than among men and younger participants. These findings are consistent with previous studies that show a greater incidence of ambivalent sexism among men than women (Becker & Wright, 2011; Glick & Fiske, 1996) and among young adults than older age groups (e.g., Hammond et al., 2018). In any case, participants show mean scores on benevolent attitudes and hostility below the scales’ average scores. This seemingly low prevalence of ambivalent sexism could be due to the profound shifts in attitudes toward women, which are becoming more favorable, likely due to the larger presence of women in public domains (European Parliament, 2021) and the growing belief that women and men are equally competent (Eagly et al., 2019). However, the data continue to reflect the disadvantaged situation of women in comparison to men in society (European Institute for Gender Equality, 2021). For this reason, we might wonder how real such low scores in explicitly stated sexist attitudes are, or whether they are attributable to other factors, such as social desirability or the self-adherence to certain stereotypical gender traits.

In this regard, as predicted, social desirability has a distinct effect in the relationship between self-ascribed gender traits and each sub-component of ambivalent sexism. In the case of benevolent sexism, our findings do not support the three-way interaction effect of feminine and masculine traits and social desirability in the expected direction. Interestingly, however, the conditional significant effect found could suggest that benevolent sexism depends on the combination of high stereotypically feminine traits and low masculine traits among individuals with low social desirability in their responses. The self-definition with feminine-stereotyped attributes, rather than with masculine ones, implies viewing women stereotypically and in limited roles and confirms benevolent sexism as an apparently flattering sexist ideology that represents women as affectionate, kind, and pure (Connelly & Heesacker, 2012). It reinforces the idea that, for women, passivity, and delicacy (being a ‘princess’) are rewarded by protection from a ‘Prince Charming’ (Mastari et al., 2019).

Contrary to what we expected, benevolent sexist attitudes increase among the participants with the abovementioned gender profile, but who provide low socially desirable responses. Our predictions on the effect of social desirability were not fulfilled, possibly because individuals did not consider it necessary to give socially desirable responses. Benevolent attitudes are so entrenched in our society (Hideg & Shen, 2019; Mastari et al., 2019) that participants do not think they need to mask their answers to gain the approval of others by giving responses that do not correspond to what they believe. Remember that people may provide socially desirable responses when they want to hide their negative attitudes (impression management) or they genuinely dislike the negative attitudes they have (Lorenzi-Cioldi & Kulich, 2015) to obtain approval from others or to avoid their disapproval (Crowne & Marlowe, 1960). If the person believes that they already have society’s approval (benevolent attitudes are viewed positively) (Chisango et al., 2015), as well as thinking that showing this benevolence will not provoke others’ disapproval, they may feel no need to give socially desirable answers. This highlights just how entrenched benevolent attitudes are in today’s society. This might be because, according to system justification theory (Jost & Banaji, 1994), it favors diffuse system justification, that is, the perception that the current social structure is fair and legitimate for women and men in different ways. Benevolent sexist attitudes may reinforce diffuse system justification by strengthening the perception that society is balanced, providing an acceptable explanation for gender inequality (i.e., women are more suited to caretaking responsibilities, while men are more suited to leadership responsibilities) and positive social identities for women and men (e.g., women are kind and men are strong). Also, benevolent sexist attitudes may strengthen diffuse system justification by emphasizing that both women and men are complementary but equal (e.g., an affectionate female nurturer complements a forceful male provider) (Connelly & Heesacker, 2012).

In the case of hostile attitudes, the moderated moderation did not appear, but the two interaction effects found take the same expected direction. Social desirability moderates the relationship between masculinity and hostile sexism. The more individuals describe themselves in a more masculine way and have low socially desirability scores, the less concerned they will be about expressing their antipathy toward women. By contrast, when individuals score high in social desirability, regardless of their self-ascribed masculinity level, they do not report hostile attitudes explicitly, presumably because of the social rejection this may generate (Chisango et al., 2015; Hideg & Shen, 2019; Mastari et al., 2019). According to system justification theory (Jost & Banaji, 1994), hostile sexism does not favor the perception of a fair society (i.e., this attitude is so openly aggressive toward women that it could not plausibly foster the perception that society is fair) (Connelly & Heesacker, 2012). This would explain why hostile sexist individuals mask or hide their open antipathy toward women. Moreover, individuals with high masculine traits, but low femininity, are those who report the highest hostility toward women, in accordance with the literature (e.g., Glick et al., 2015; Spence & Buckner, 2000).

This study extends our knowledge of the relationships between self-adherence to gender traits, ambivalent sexism, and social desirability. Firstly, in general, our results showed that self-adherence to gender traits is linked to ambivalent sexism, confirming results from previous studies (e.g., Glick et al., 2015; Lameiras et al., 2007; Spence & Buckner, 2000). Secondly, we contribute to the literature by explicitly examining the influence of social desirability in the association between gender-typed personal traits and ambivalent sexism. Our findings reveal that socially desirable responding seems to play a more robust role in masking hostile sexism than benevolent sexism, and system justification theory (Jost & Banaji, 1994) helps us to better understand the results. Benevolent sexism encourages the perception that the system is fair, legitimate, and balanced, whereas hostile sexism has precisely the opposite effect.

The fact that benevolent attitudes were not acknowledged as sexist and were perceived more positively prevents women from reacting and protecting themselves against benevolent sexist behaviors (Dardenne et al., 2007). Thus, as benevolent sexist attitudes are seen as more socially acceptable, they are more difficult to eradicate. Indeed, a growing body of research now suggests that the consequences of benevolent sexism may be even more damaging than those resulting from explicitly hostile sexism. Benevolent sexist attitudes are associated with attitudes that tolerate domestic violence, with sexual harassment, with disapproving responses to rape victims (see Rollero & Fedi, 2012) and even with appointing women to more unstable positions in which they are more likely to fail than those occupied by men (i.e., the glass cliff phenomenon) (see Ryan et al., 2016).

Thirdly, strategies must be developed to weaken the adverse consequences sexism has on women in all spheres of life. Obviously, this is imperative in the case of hostile sexism, but it is also crucial for benevolent sexism, since, as we mentioned previously, this type of sexism is pervasive and may not be viewed as prejudice. Barreto and Ellemers (2005, p. 633) argue that “prejudice first needs to be perceived before it can be challenged as an illegitimate cause of social inequality.” Strategies aimed at increasing perceptions of benevolent sexism as a prejudice that threatens gender equality are therefore a necessary previous step to prevent such sexism from “undermining women’s motivation to engage in direct action to improve their lower status” (Becker & Wright, 2011, p. 74). In this line, Becker and Swim (2011) used sexism diaries to reduce women’s validation of benevolent sexism by heightening their sensitivity to sexist attitudes in their lives. This intervention reduced women’s benevolent attitudes, but the men’s attitudes remained unaffected. Another intervention used reading, watching videos and reflective writing to successfully raise students’ consciousness of heterosexual and male privilege (Case et al., 2014). A training program on gender designed for female and male participants that incorporated evidence about gender as a social construction, resources, and social obligations and ideologies, was also effective in reducing hostile and benevolent sexism (de Lemus et al., 2014).

Conclusion

In conclusion, gender stereotypes (including those we use to self-describe) continue to be pervasive among all members of society and their effect is reflected in sexist attitudes, despite legislation and gender equality policies aimed at reducing sexism. The self-adherence to stereotypical gender traits and social desirability are connected to benevolent and hostile sexism, although in distinct ways. The profile of those with benevolent attitudes (i.e., participants who self-attribute mostly feminine traits) is different from those with hostile attitudes (i.e., participants who self-ascribe mainly masculine traits). In addition, the need to gain approval from others or––more importantly––avoid their disapproval, is more manifest in the case of hostile sexism (i.e., individuals tend to offer socially desirable responses that mask their hostility toward women), whereas there is less urgency to hide benevolent attitudes. Unlike with hostile sexism, the shared perception that the status quo is fair for both men and women encourages the normalization, acceptance, and rewarding of benevolent sexism in our society. However, the pervasiveness of such sexist attitudes in society and their apparent benefits for women constitutes a serious threat, since it inhibits collective action for social change to eradicate benevolent sexism and promote equality between women and men.

Our findings shed light on the relationships among subjects’ self-adherence to stereotypical gender traits, ambivalent sexist attitudes, and social desirability, without drawing conclusions about the causality of the associations. However, several considerations could be addressed in future research. Firstly, the use of college student participants in the study may limit the generalizability of our results to other settings, although it should be noted that a significant body of research on sexism has been carried out among young people (e.g., adolescents, undergraduate students) (e.g., Mastari et al., 2019). In addition, the results from the BSRI should be interpreted with caution, since perceiving oneself as high in stereotypically masculine traits does not always imply that one has a ‘masculine identity,’ nor does a self-perception high in stereotypically feminine traits necessarily imply having a ‘feminine identity’ (Glick et al., 2015). For this reason, future studies could use specific instruments designed to measure masculine and feminine identities (e.g., The Masculine and Feminine Self-Disclosure Scale, Snell, 2013). In addition, in our study social desirability seems to have a floor effect; that is, responses are mostly socially desirable, with a low variance, which may be influencing the significance of the interactions. For this reason, the effect of social desirability in sexism could be further explored through other measures (e.g., Modern Sexism Scale, Swim et al., 1995; Neosexism Scale, Tougas et al., 1995).