{PSOCIAL}

Galtung (2016) states that it is difficult to speak of a violent culture, since most of its elements would need to give way to violent behaviors and beliefs. However, a first level of analysis for the understanding of sexual violence accounts for a cultural component that comprises specific types of violence, expressed in prejudiced attitudes, legitimized through conservative ideological frameworks. In this regard, the RWA helps to understand the perception towards authority figures and the search for the preservation of social order (Altemeyer, 1996; Funke, 2005), and the SDO refers to the predisposition of individuals towards the maintenance of hierarchical and non-egalitarian intergroup relations (Henríquez et al., 2020; Pratto et al., 1994). Both, despite their distinct motivations, represent ideological dimensions that prescribe how society should be structured and act as good predictors of prejudice and social attitudes that justify violence (Duckitt & Fisher, 2003).

Gender roles are traditionally defined as beliefs that society assigns to women and men regarding the performance of both in differentiated tasks (Páez et al., 2004; Zubieta et al., 2011); while gender stereotypes are rigid categories, about physical and behavioral characteristics, differentially attributed to women and men (Páez et al., 2004). Thus, in traditional societies, women are attributed a strictly and consistently "feminine" role, while men are attributed a strictly "masculine" role, questioning the possibility that women can assume behavioral patterns that are characteristically masculine, or vice versa (Restrepo & Aponte, 2009).

From these gender roles follow: (1) the descriptive component, referring to what a member of society does, in other words, the stereotype; and (2) the prescriptive component, referring to what society determines for each member to do and which, in turn, reinforces the stereotype (Zubieta et al., 2011). Both components, established as social norms, limit the behavior of men and women and the possibility of varying their behaviors; therefore, those who transgress the prescriptions established by the social power structure are sanctioned (Zubieta et al., 2011; Gaunt, 2013). In this sense, elements such as language, clothing, work, family relationships, among others, provide the parameters to delineate behavior in accordance with the assigned gender (González, 2008).

Consistent with the above, some stereotyped roles about masculinity and femininity converge in sexism, defined as a set of traditional beliefs about the characteristics and behaviors considered appropriate for women and men, which have a pernicious influence on women by perpetuating their subordination to men (Glick & Fiske, 1996, 2001; Glick et al., 2000). Consequently, sexism is expressed in both subtle and hostile forms of exclusion. Under this view, sexism is no longer understood as a uniform antipathy towards a certain sex, but is interpreted in a deeply ambiguous form that is expressed by the coexistence between feelings of sympathy and hostility (Cárdenas et al., 2010).

The concept of ambivalent sexism is then introduced, which would be made up of two components: Benevolent Sexism [BS] and Hostile Sexism [HS] (Arbach et al., 2019). The BS gives women the status of a fragile person in need of protection by men (Cárdenas et al., 2010), while the HS refers to attitudes with violent connotations towards women, based on the supposed female inferiority and sexual power exercised by women over men, associating them with threat and manipulation (García-Leiva et al., 2007). Both types of sexism are intended to legitimize and reinforce the subordinate position of women, in other words, gender inequality (Zubieta et al., 2011).

Previous studies have revealed a linear association between sexism, favorable attitudes towards verbal and psychological violence (García-Leiva et al., 2007), patriarchal beliefs, and symbolic violence (Pecho, 2017). It is important to emphasize that BS is also associated with the acceptance of myths about sexual violence (Janos & Espinosa, 2018) and the attribution of blame to the victim when facing this type of violence (Abrams et al., 2003). Meanwhile, the HS reinforces beliefs and favorable attitudes towards violence against women constituting one of the sociocultural risk factors leading to the existence of this violence (Heise & García-Moreno, 2003).

Both expressions of sexism are associated with ideological measures of RWA and SDO in Peruvian samples. Specifically, it is often observed that the magnitude of the RWA effect tends to be higher in the BS, considering that the RWA reinforces the preservation of a traditional social order; while the magnitude of the SDO effect tends to be higher in the HS, considering that it emphasizes the dominance, even by force, of some groups over others (Janos & Espinosa, 2018; Rottenbacher et al., 2011).

Peruvian society is characterized by being mostly chauvinistic and patriarchal (Fernández & Duarte, 2006); stereotyping women as vulnerable, passive and dependent on men; thus, women whose characteristics escape from the stereotypical representations will be portrayed as transgressors of their traditional role, a fact that will turn them into inciters of violence and responsible for any negative consequence against them (Janos & Espinosa, 2015; 2019). In that line, at the political, legal and social level, this patriarchal position hinders the recognition and criminalization of acts of sexual violence (Janos & Espinosa, 2019; Velázquez, 2003).

This is evidenced in the study of Espinoza (2000), where it was found that 90.9% of women identified rape as the only sexual violence expression, while a minority of respondents related sexual violence to other forms of aggression such as obscene gestures or attempted rape. Likewise, more recent studies show that, of the 78% of denunciations received by telephone referring to some type of violence against women, only 8.7% were considered as sexual violence, maintaining almost the same percentage distribution in terms of the statistics of cases reported in police stations and women's emergency centers (Unidad de Generación de Información y Gestión del Conocimiento [UGIGC], 2019).

Public institutions can incite and perpetuate dynamics of inequality in society, consequently, it is important to examine the representations of inclusion at the institutional level (Sperber, 1990). Likewise, the determination of the degree of violence is often evidenced in an institutional and collective level (Cohen & Nisbett, 1997), so sexual violence can also occur in institutions that should be considered "safe" (Ariza, 2013; Organización Mundial de la Salud [OMS], 2013). In this sense, this research considers as a priority to analyze how sexual violence is represented in formative contexts of police cadets and university students.

Specifically, Borja (2019) identifies that in police trainee there is a representation that reinforces, on sexist grounds, the responsibility of women when they are victims of sexual violence. Consistent with these findings, the police have become the propitious scenario to vulnerate some rights, especially when dealing with women rights (Defensoría del Pueblo, 2019). A finding of concern made by the Defensoría del Pueblo (2019) is that some acts of violence and discrimination are accepted and normalized in these institutions, which combined with the deficient preparation and functional specialization process in the training of police officers (Ramos, 2017) could explain the insufficient capacity of these institutions to effectively address issues of family and sexual violence (Defensoría del Pueblo, 2017).

In contrast to police training, the Ley N° 30220 of 2014 states that the university implies an academic community oriented towards research and teaching, whose formation should allow for a clear awareness of the social reality of the country and encourage the rejection of all forms of violence. Sexual violence, however, is often invisibilized and complicates the recognition of patterns of violent behavior of a sexual nature among members of university communities (Janos & Espinosa, 2015; Póo & Vizcarra 2008; Vizcarra & Póo, 2011).

Considering the information above, the following research question arises: What is the relationship between ideology, sexism and beliefs about sexual violence among police trainees and university students? In that sense, the general objective is to establish a comparative view of the relationships between ideology, ambivalent sexism and beliefs about sexual violence among police trainees and university students. In addition, the first specific objective is to compare the aforementioned variables between the groups of police trainees and university students. As a second specific objective, it is proposed to develop an instrument that assesses beliefs about sexual violence against women.

With regard to the study hypotheses, the following is proposed: H1: There are significant differences on ideology (RWA and SDO), ambivalent sexism (BS and HS), and sexual violence beliefs among university students and police trainees. Lower levels of conservative ideology, sexism and a broader representation of beliefs about sexual violence are expected among university students because, despite deficiencies in the educational system, they are exposed to more critical information about violence. H2: There is a positive relationship between ideology (RWA and SDO), ambivalent sexism (BS and HS) and sexual violence beliefs in the general sample of the present study (college students and police trainees).

The sample consisted of a total of 277 participants, 196 university students (women = 55.6% and men = 44.4%) and 81 police trainees from an officer school of the Peruvian National Police (men = 100%). The age of the participants was between 18 and 25 years (M=21.5, SD=.49).

Regarding ethical considerations, an informed consent form was prepared clarifying the free and voluntary participation, the strict confidentiality of the information including the researchers' data, the objectives of the research, and informing the participants about support contacts in case the instrument mobilized feelings and emotions that required professional assistance through a referral booklet.

In addition, the instruments for measuring the constructs were reviewed, the scale of beliefs about sexual violence was constructed and coordination was made with the pertinent authorities to gain access to the institutions chosen as part of the sample. It is pertinent to mention that the tests used were applied through online formats in the case of university students and printed in the case of the cadet sample.

IBM SPSS 25 Statistics and R-studio software were used for the analysis and interpretation of data from the present research. During the first phase of the analysis, evidence of validity and reliability of the constructed instrument was collected. For the collection of evidence of internal structure validity, an exploratory factor analysis was performed to identify and analyze the respective dimensions of the construct. Thus, the items that did not contribute to the best expression of the construct were identified. Mardia's coefficient was used for multivariate normality.

The extraction method used was generalized least squares (GLS), the number of rotations was set to 250 and the statistical criteria for the relevance of the EFA were the Kaiser-Meyer-Olkin (KMO) measures, greater than 0.70 and Barlett's test of sphericity (p < 0.05). Along these lines, the criteria for determining the number of factors are 3: Kaiser, Horn, MAPS.

Based on the results obtained from the EFA, we proceeded to eliminate those items that were not relevant to the scale. The criteria for eliminating items were having a negative factor load and/or close to zero. Regarding the reliability of the test for each factor, it was verified that the Cronbach's alpha obtained was greater than 0.5 (Kaiser, 1974) to be considered acceptable. Similarly, internal consistency was checked through corrected item-test correlation values greater than 0.3. Then, descriptive analyses of the constructs studied were performed and statistical contrasts were made between the groups of participants using the Student's t-statistic. Subsequently, the analysis of correlations between the variables per group was carried out using Pearson's correlation coefficient. Finally, a path analysis was performed to analyze the nature and magnitude of the relationships between the variables studied using an integrated model.

To determine the dimensions of the sexual violence belief scale, an exploratory factor analysis (EFA) over 37 items was performed. Mardia's test for multivariate normality analysis ("Mardia's test for multivariate normality") was performed and, in the absence of normality, proceeded with the generalized least squares extraction method and varimax rotation.

The first AFE obtained a good factorial adjustment (KMO = .945) and an adequate value for the Kaiser-Meyer-Olkin measure, greater than .60 (Kaiser, 1974), with a significance of p<.05 in Bartlett's test of sphericity. A first six-component model was obtained, explaining 62.16% of the total variance, with factor loadings from .462 to .798.

Then, 6 items with high factor loadings in more than one factor were eliminated (crossloading). In a subsequent processing, 11 items were eliminated whose factorial grouping did not correspond to a component consistent with the theoretical proposal and the objective of the instrument. Finally, a new three-component model was obtained (KMO=.933, p<.05 in Bartlett's test of sphericity) explaining 60.47% of the total variance with factor loadings between .813 and .584 and correct fit (see Table 1). These three components were labeled: (1) Subordination of women, which encompasses beliefs associated with women's willingness and availability to satisfy men's desires, (2)Exaggeration of women, a characteristic attributed when faced with any manifestation of denunciation or accusation of violation of their sexual rights, devaluing and discrediting such accusation, and (3) Attribution of blame to women, a belief group that holds women responsible for the sexual violence experienced by them by appealing to different causes such as behavior, dress or attitude. All dimensions obtained adequate reliability indicators.

In response to the main objective and the first hypothesis, statistical tests to determine if there were significant differences between students from an officer school and a university were conducted. Thus, in relation to ideology, police trainees showed statistically significant higher scores on RWA, while university students scored higher on SDO, although this difference was only reported at the descriptive level.

On the Violence Against Women Inventory statistically significant differences are found only on the Subordination of Women dimension, university students showed significantly higher scores compared to police trainees. Likewise, at the descriptive level, police trainees score higher on the Blame and Exaggeration dimensions (See Table 2).

In reference to the specific objective, the significant correlations between the variables SDO, RWA, SB, SH and Beliefs about sexual violence in the general sample are detailed.

Specifically, the level of right wing authoritarianism correlates positively with the indicators of BS (r= .37; p< .05) and HS (r= .19; p< .05). Also, a positive relationship is seen with two of the sexual violence belief factors: Attribution of blame to the woman (r= .22; p< .05) and Exaggeration of the woman (r= .21; p< .05). Regarding the level of social dominance, this variable correlates positively with BS (r= .32; p< .05) and HS (r= .54; p< .05). In addition, a positive relationship is seen with the three factors of beliefs about sexual violence.

With respect to the level of BS, this correlates positively with HS (r= .69; p< .05). Also, a positive relationship is seen with the factors of beliefs about sexual violence: Subordination of the woman (r= .41; p< .05), Attribution of blame to the woman (r= .47; p< .05) and Exaggeration of the woman (r= .45; p< .05). In the same line, the level of HS, correlates positively with the factors of beliefs about sexual violence: Subordination of the woman (r= .63; p< .05), Attribution of blame to the woman (r= .56; p< .05) and Exaggeration of the woman (r= .59; p< .05). On the other hand, Subordination of women correlates positively with Attribution of blame to women (r= .59; p< .05) and Exaggeration of women (r= .71; p< .05), while Attribution of blame to women correlates positively with Exaggeration of women (r= .65; p< .05).

At the group level, in police cadets there are significant and inverse correlations between the variables RWA with SDO and RWA with Subordination of women. On the other hand, significant and direct correlations were found between SDO with the factors of the Sexism and Beliefs dimensions, BS with HS and the Beliefs factors. Likewise, HS presented a significant and direct correlation with the three Beliefs dimensions, and likewise, Subordination of women with Attribution of blame to women and Exaggeration of women, and Attribution of blame to women with Exaggeration of women.

As for the university students, the RWA variable presented a significant and direct correlation with the factors of the Sexism dimension and with those of the Beliefs dimension. The same results were found for the SDO variables, the Sexism and Beliefs factors.

In order to comprehensively analyze the nature and magnitude of the relationships between the variables studied, a path analysis was performed. A hypothetical model of the relationships between variables was proposed through a path analysis that presents three levels (see Figure I). At the first level, the RWA and SDO ideology variables are considered as exogenous, structural and stable with direct influence on the other variables, placing sexism, Benevolent and Hostile at a second endogenous level. Beliefs about sexual violence are considered at a third level, as an output cognitive element that would be affected by both ideology and sexism.

The criteria for evaluating the goodness of fit of the proposed models to the data used were the following indices: the quotient resulting from the division of the Chi-square statistic (c2) and the degrees of freedom of the model (c2/gl), the Bentler-Bonett Comparative Index (CFI), the Normalized Bentler-Bonett Index (NFI), the Steiger-Lind root mean square error of approximation (RMSEA) and the Standardized Root Mean Square Residual (SRMR). According to the criteria of Ruiz, Pardo and San Martín (2010) the fit values considered acceptable were the following: c2/gl ≤ 3, CFI ≥ 0.95, NFI ≥ 0.90 and RMSEA ≤ 0.08. In addition, the SRMR criterion is 0.08.

The goodness-of-fit indicators for hypothetical model 1 are generally good: χ2/gl =.192 CFI= 1, NFI=1, RMSEA= .0001 and SRMR=.00. In particular, both RWA and SDO have been found to retain a direct and highly significant relationship with the variables BS and HS, albeit with medium-low effects. In turn, these variables are directly related to Subordination of women, Attribution of blame to women and Exaggeration of women, but only HS was significant with a medium effect size.

Likewise, RWA has presented a significant and inverse influence on Women's Subordination (β=-0.107, p<0.05), as well as a significant and direct relationship between SDO and Women's Subordination (β=.125, p<0.05), both variables have a medium-low effect magnitude on Women's Subordination. On the other hand, neither RWA nor SDO were significant as exogenous variables influencing Women's Blame Attribution and Women's Exaggeration, the same occurred with BS, a variable that did not present significant influence on Women's Subordination and Women's Exaggeration.

Based on the results described above, an alternative model with three analytical levels is proposed (see Figure 2). The first level comprises the ideology variables, at a second level the sexism variables and at the third level the sexual violence belief variables. The particularity of the alternative model is that only the significant regression lines are kept in the analysis, thus achieving a more parsimonious model of relationships.

The goodness-of-fit indicators of the alternative model are an improvement over the indicators of the hypothetical model 1 (in most cases): χ2/gl =1.532, CFI= .995, NFI=.986, RMSEA= .044, SRMR=.03. Also, further analysis of this model suggests that the proposed statistical relationships in all cases are significant and highly significant, which would allow a more accurate assessment of the effects of RWA, SDO, HS and BS on Women's Subordination, Women's Blame Attribution and Women's Exaggeration. Thus, RWA and SDO retain a direct and medium-low effect relationship on HS and BS. In turn, HS continues to influence with a medium to medium-high magnitude on Subordination of the woman (β=.606, p<.01), Attribution of blame to the woman (β=.476, p<.01) and Exaggeration of the woman (β=.597, p<.01); while BS only influence Attribution of blame to the woman (β=.122, p<.05). On the other hand, SDO still directly influences Women's Subordination (β=.088, p<.05) and RWA still retains an inverse relationship with Women's Subordination (β=-.169, p<.01).