This is the accepted version of the journal article: Gómez-Casillas, Amalia; Van Damme, Maike; Permanyer, Iñaki. «Women’s and men’s status: revisiting the relationship between gender equality and intimate partner violence against women in Europe». Journal of Interpersonal Violence, 2023. DOI 10.1177/08862605231158760 This version is available at https://ddd.uab.cat/record/275337 under the terms of the license Artículo de revista: Gómez-Casillas, Amalia; van Damme, Maike; Permanyer, Iñaki (2023) ‘Women’s and men’s status: revisiting the relationship between gender equality and intimate partner violence against women in Europe’. Journal of Interpersonal Violence: 088626052311587 (ISSN 1552-6518) DOI: 10.1177/08862605231158760 Women’s and men’s status: Revisiting the relationship between gender equality and intimate partner violence against women in Europe. Amalia Gómez-Casillas Ph.D.1 Centre d’Estudis Demogràfics, (CED-CERCA), UAB. ORCID: 0000-0002-3528-7844 Maike van Damme PhD Centre d’Estudis Demogràfics, (CED-CERCA), UAB. ORCID: 0000-0003-1080-9789 Iñaki Permanyer Centre d’Estudis Demogràfics, (CED-CERCA), UAB. ORCID: 0000-0002-7051-5144 Publication citation: Gómez-Casillas, A.; van Damme, M.; & Permanyer, I. (2023) “Women’s and men’s status: Revisiting the relationship between gender equality and Intimate Partner Violence against Women in Europe”. Journal of Interpersonal Violence. https://doi.org/10.1177/08862605231158760 Abstract The first European Union Survey on Violence against Women (EU-VAW) released in 2014 revealed the unexpected result indicating that the world’s most egalitarian countries have relatively high rates of Intimate Partner Violence Against Women (IPVAW). This phenomenon, referred to as the ‘Nordic Paradox’, revived a heated, intermittently ongoing discussion dating back four decades where several competing hypotheses about the relationship between gender inequality and IPVAW have been proposed, but no consensus has been reached. The main aim of this paper is to revisit the most important of such hypotheses proposed in the last four decades, while proposing a new one that could potentially throw some light on understanding the ‘Nordic Paradox’. Multilevel linear regression models are estimated using data from the EU-VAW survey conducted in 2012, and an alternative operationalisation of the Gender Equality Index (our measure of gender equality). We did not find any significant effect of gender equality on IPVAW 1 Corresponding author: Amalia Gómez-Casillas PhD – agomez@ced.uab.es, +34 935 81 30 60 – Centre d’Estudis Demogràfics, Carrer de Ca n’Altayó, Edifici E-2, Campus de la UAB, 08193 Cerdanyola del Vallès, Spain 1 repetition. However, we found that higher country-level status of women and men go together with less IPVAW, with a larger effect of women’s status in economic domains compared to the impact of men’s economic status, and a larger effect of men’s overall status. These findings support the Marxist feminist hypothesis, stating that women’s absolute status in the economic and labour domain is critical in lessening IPVAW, as women’s real and potential access to resources is key for leaving a violent relationship. At the same time, our results support the ‘male privilege protection’ hypothesis, which states that gains in women’s status in certain domains –such as in the economic sphere considering our results for the European Union– would not suppose a threat to men, allowing ameliorative effects. In contrast, if the overall status of men is threatened, backlash effects would be triggered. Introduction The first comparable survey among the European Union Member States, conducted in 2012 the European Union Agency for Fundamental Rights (FRA, 2014), indicated that the most gender- egalitarian countries in Europe –Denmark, Finland, and Sweden– were among the countries with a higher prevalence of Intimate Partner Violence Against Women (IPVAW). These surprising findings, referred to as the ‘Nordic Paradox’ (Gracia & Merlo, 2016), have revived a heated and intermittent discussion dating back four decades. On the two opposite ends of this debate are the liberal feminists and the radical feminists, representing the ameliorative and the backlash hypothesis respectively. While many scholars have addressed this debate directly (Ivert et al., 2019; Gracia et al., 2019; Permanyer, & Gómez-Casillas, 2020; Humbert et al., 2021) or indirectly (Sanz-Barbero et al., 2018), other equally interesting debates have been side-lined. One of those important debates involves a confrontation between Marxist feminist theoretical claims on the one hand and on the other hand the positions of liberal and radical feminists. Marxist feminists claim that women's absolute status is a better predictor of IPVAW than the gender gap –namely the relative position of women compared to men’s–, stressing in particular women's economic status as the key aspect leading to a reduction of IPVAW. On the other side of this debate, liberal and radical feminists argue that women’s status relative to men’s status (i.e., the gender gap) is the key aspect impacting IPVAW. Surprisingly, men’s absolute status –men’s status per se, not comparing their status to women’s– has been virtually absent ‘Nordic Paradox’ debate, despite the fact that male partners are the main perpetrators of violence against women (see for example Stöckl, 2013). The main aim of this paper is to test four existing competing hypotheses: the ameliorative hypothesis proposed by liberal feminist scholars, the backlash hypothesis proposed by radical 2 feminist academics, the Marxist feminist’s hypothesis –also known as the economic marginalization hypothesis–, and finally, the new ‘male privilege protection’ hypothesis. This last hypothesis is new and states that women are ‘allowed’ to increase their status to a certain extent, but only if men do not feel that their own status is threatened. Hence, this hypothesis could potentially explain ameliorative and backlash effects empirically observed. To test these hypotheses, we introduce a couple of methodological improvements. First, we use an IPVAW indicator that is sensitive to the repetition of violence against a woman. In addition, this indicator accounts for the violence that is occurring in the present, and thus is perpetrated by the current partner. Thus, our dependent variable is the repeated violence against women perpetrated by the current partner. Second, we separate the Gender Equality Index (GEI) created by the European Institute for Gender Equality (EIGE) into two different indices: one that operationalises the gender gap (the relative status of women compared to men) and one that measures overall population’s achievement or attainment (we will refer to them as synonyms throughout the text). The method we use is multilevel regression analysis using data from the European Union’s Violence Against Women survey (EU-VAW survey hereafter) conducted by the European Union Agency for Fundamental Rights (FRA, 2014). Background: Old and new hypotheses Previous seminal theoretical contributions to the gender-based violence field of studies have generally understood violence against women as an expression of gender inequality (Daly, 1978; Brownmiller, 1975; Dobash & Dobash, 1979), and have thus questioned whether violence against women would increase or decrease when higher levels of gender equality have been achieved. In this debate, there are two major positions: on the one hand, liberal feminists mostly support the ameliorative hypothesis, which contends that increased levels of gender equality are associated with less violence against women. On the other hand, radical feminists mostly support the backlash hypothesis, which argues that men become more violent mainly in order to regain the power they have lost due to higher levels of gender equality (Yllö, 1984). Although this backlash hypothesis is often presented as being the theoretical extreme opposite of the ameliorative hypothesis, some scholars who study IPVAW tend to embrace more nuanced positions as well. While some authors have suggested that the backlash is a temporary effect that is leading to a sustained ameliorative situation (Brownmiller, 1975; Russell, 1975), others (Whaley, 2013) have suggested that there is an interplay of backlash and ameliorative processes, which also aligns to Stamatel’s (2018) findings. 3 While the debate on ameliorative versus backlash effects is well-recognised and well-addressed in empirical studies, the debate on the impact of women’s relative versus absolute status is less well-known. While liberal feminist scholars provide support to the ameliorative hypothesis, and radical feminists to the backlash hypothesis, they both agree on pointing out that women’s status relative to men’s status –gender inequality– is the key factor explaining gender-based violence (Vieraitis, Britto, & Morris, 2015 provide a thorough explanation). By contrast, Marxist feminist theory posits that increases in women’s absolute status and not in their status relative to men, and more specifically, women’s access to economic resources, will reduce rates of violence against women. Following Jaggar (1983), in the capitalism the structural position of women is subject to specific forms of oppression derived from their exclusion from wage labour and confinement in the domestic sphere doing housework and childbearing. Therefore, by being excluded from labour and access to resources, women are oppressed by men from a structural standpoint, thus making them targets of male violence. This hypothesis is often referred to as the women’s economic marginalization hypothesis. Interestingly, while men’s violent reactions have been implicitly or explicitly considered in relation to women’s gains in relative or absolute status, the role of men’s status in violent victimisation has seldom been surveyed or examined in Europe. Men have been viewed as a ‘gendered social group’ starting with the early feminist contributions (i.e., Brownmiller, 1975, Yllö, 1984). Thus, it has been assumed that a man’s individual behaviour is shaped by his structural position as a member of a group, in line with social class theory. However, the idea that men should be considered as a monolithic social group has been contested by authors such as Connell (Connell, & Messerschmidt, 2005), who proposed approaching this issue from a dualistic perspective, whereby hegemonic masculinities are conceived in opposition to non-hegemonic ones. Conversely, other authors have pointed out that even if not all men are connected with the institutions of male dominance, most benefit from these institutions (Donaldson, 1993). These scholars therefore argued that instead of applying a dualistic perspective to this issue, hegemonic masculinity should be considered as a bloc where non-hegemonic masculinities are absorbed and reconfigured to ensure the reproduction of hegemonic masculinity (Demetriou, 2001). In sum, these authors contended that all men benefit from the ‘patriarchal dividend’ (as proposed by Connell), even though its intragroup distribution is also hierarchically structured. According to Kosakowska-Berezecka et al. (2020) men withdraw their support for gender equality when their status is threatened, which is aligned with the backlash hypothesis. The backlash hypothesis is often presented as referring to a phase in which men get used to equality. Therefore, the way to gender equality is not linear and depends on the interlock of structural and other forms of status, which involves interplays of backlash and ameliorative processes where men, as a social group, would bit by bit get used to equality, with at times resisting the process 4 towards this end. Kosakowska-Berezecka et al. (2020) propose to understand gains in gender equality by mirroring the zero-sum belief claiming that a group’s gain is at the expense of another group’s losses (Esses, Dovidio, Jackson, & Armstrong, 2001). There would be sources of status bargaining where women are ‘allowed’ to status acquisition in certain areas in an amelioration process while in other areas backlash effects would be triggered. Similarly to Marxist feminist positions, structural aspects –such as income and labour domains– would be the most sensitive issues, as these are limited resources available to all the members of society. Moreover, gender social groups would play a role in protecting these resources for their own gender class, thus producing a feedback process that reaffirms the own group status. Data We use data from the FRA’s EU-VAW survey (FRA, 2015) of women who reported having a heterosexual orientation as a strategy to select those couples in which the partner is male, a key aspect to pursue our goal of understanding how women’s and men’s status shapes intimate partner violence victimisation. Additionally, these women were currently married, cohabitating, or involved in a relationship without cohabitation. The initial sample of heterosexual women with a partner is 30,284. The total analytical sample consists of 27,195 women (about 10% of all women were dropped because of missing values on one or more of the variables to be analysed). Methods Methodological improvements We improve previous research methodologically in two ways. First, we use an improved indicator of gender inequality: namely, gender gaps. Most of the studies that have looked at the ‘Nordic Paradox’ debate have used the EIGE’s Gender Equality Index (GEI) to operationalise gender equality (Ivert et al., 2019; Sanz-Barbero et al., 2018; Permanyer & Gómez-Casillas, 2020; Humbert et al., 2021). The GEI is a composite indicator providing a unique score representing the overall achievement in six gender-relevant core domains: ‘work’, ‘money’, ‘knowledge’, ‘time’, ‘power’, and ‘health’. (see EIGE, 2017: Annex 1). While this index has led to improvements in the availability of gendered data in the European Union, one of its limitations is that high index scores reflect both small gender gaps and high levels of overall achievement of both women and men (i.e., population achievement). To overcome this limitation, Permanyer (2015) has proposed getting rid of this overall population attainment component, and instead creating a Gender Gap Measure that should provide less misleading gender inequality estimates. Therefore, the EIGE’s GEI can be decomposed into two components: the Gender Gap Measure 5 that estimates women’s attainment relative to that of men, and an Overall Population Attainment Measure that reflects average achievement, without accounting for gender differences. Moreover, we use an IPVAW indicator that is sensitive to the frequency of victimisation to overcome the limitations of the mainstream IPVAW prevalence indicator. As prevalence indicators do not differentiate between chronic and episodic violence, it is necessary to introduce an indicator of repetition, which is a continuous measure accounting for the frequency of IPVAW. Many studies have highlighted the key role of repetition in portraying victimisation (Walby & Towers, 2017; Walby et al; 2017; Walby, Towers & Francis, 2016). The close relationship between partners creates the conditions for chronic victimisation, such as repetition of violent episodes. Thus, our repetition-sensitive indicator allows us to differentiate between episodic and chronic violence (Gómez-Casillas, 2018; Permanyer & Gómez-Casillas, 2020). We also follow previous contributions (Gómez-Casillas, 2018; Permanyer & Gómez-Casillas, 2020) that argued that it is conceptually relevant to differentiate IPVAW that occurred in the past and was perpetrated by the previous partner from IPVAW that was perpetrated by the current partner. Although information about violence perpetrated by previous partners is also important to examine (and has a higher prevalence than the violence committed by the current partner), merging present and past victimisation could lead to misleading conclusions (Permanyer & Gómez-Casillas, 2020). Moreover, the survey does not allow us to account for the number of previous partners nor consider male partners’ individual-related status (i.e. partner’s educational attainment or employment status), which deems critical to our research objectives. In sum, higher prevalence measured using traditional IPVAW indicators (i.e., perpetrated by any current or former partner) could also mean that union formation and dissolution occur more often, that a woman’s probabilities of escaping a violent relationship are higher, or that a man’s probabilities of reoffending against a new partner are higher (Permanyer & Gómez-Casillas, 2020). As a consequence, by using a repetition-sensitive indicator, we account for violence chronicity; and by taking into account IPVAW perpetrated by the current partner only, we accurately assess violent victimisation occurring in the present. Dependent variable Due to the abovementioned conceptual reasons, the dependent variable is the repetition-sensitive indicator measuring physical or/and sexual violence perpetrated by the current partner, which was introduced in a previous study (Permanyer & Gómez-Casillas, 2020). The survey asked women if they were experiencing specific forms of violent victimisation using a modified version of the Conflict Tactic Scale; and, if the respondents said they were, they were asked if they had 6 experienced violence once, two to five times, or six or more times (see Table 1). The thusly constructed repetition-sensitive indicator had values ranging from 0 to 100; with 0 indicating no violent victimisation and 100 indicating that the woman had experienced every form of violence with the highest frequency. This indicator is therefore a slightly modified version of Permanyer & Gómez-Casillas’ (2020) proposed indicator with values ranging from 0 to 1 (this is merely a change in the scale of the indicator). Independent variables at the contextual level While some early studies focused on women’s dependency on their male partners at an interpersonal level (i.e., Kalmuss & Straus, 1982), Yllö (1984) enriched the discussion by highlighting that women and men bring resources to their couple relationships as members of a ‘gender class’. Yllö (1984) uses this terminology to acknowledge that she refers to structural inequalities, which implies differences in the access to resources as a result of institutionalized inequalities, rather than as individuals, which places these studies in a wider social context, considering on the one hand, status at a contextual level and, on the other hand status at an individual level. On the contextual level, we include five indicators: (1) the original GEI; (2) our modification of the index to measure the Gender Gap Measure; (3) our modification of the index that captures only the achievement component (Overall Population Attainment); (4) women’s achievement and men’s achievement separately; and (5) women’s and men’s achievement in the work and money domains. All of these constructed indicators are derived from the European Institute for Gender Equality’s Gender Statistics Database, provided by the EIGE in response to our request. We describe each of the indicators in detail below (the formulas for their construction are available in Annex 1). Gender Equality Index (GEI) First, we assess the impact of the EIGE’s Gender Equality Index (EIGE, 2021a) –a summary measure of 31 indicators for six domains ‘work’, ‘money’, ‘knowledge’, ‘time’, ‘power’, and ‘health’)– on violent victimisation using the indicator as proposed by the EIGE, but slightly changing it to have values ranging from 0 to 1, with 0 indicating full inequality, and 1 indicating full equality. Gender Gap Measure 7 Following Permanyer’s (2015) contributions, we decompose the EIGE’s GEI into two indicators, a Gender Gap Measure and an Overall Population Attainment Measure, which are based on the 31 input indicators, corresponding to the six domains (‘work’, ‘money’, ‘knowledge’, ‘time’, ‘power’, and ‘health’) and used by the EIGE for their GEI index. With respect to the Gender Gap Measure, we follow Permanyer’s (2015) methodology, but apply his approach to the list of indicators for 2012 (see EIGE, 2017: Annex 1). This newly created index has values ranging from 0 to 1, with 0 indicating full inequality, and 1 indicating full equality. Overall Population Attainment Measure We also generate an Overall Population Attainment Measure that focuses on the average attainment for women and men together. This measure considers each average indicator score for women and men, as provided by the EIGE’s Gender Database. The process of turning the Gender Equality Index into an Overall Population Attainment Measure can be challenging because certain indicators are not relevant for overall achievement, or because it is not possible to provide adequate measures (i.e., in the domain of power, the indicator includes shares of female/male representatives; more details are provided in Annex 2). The Overall Population Attainment Measure has values ranging from 0 to 1, with 0 indicating no achievement for the indicators of the GEI index (i.e., in the six domains of (work, money, knowledge, time, power and health), and 1 indicating full achievement. Women’s and men’s absolute (economic) achievement To assess the Marxist feminist approach, we construct the achievement component separately for women (again, with values ranging from 0 (no achievement) to 1 (full achievement), as well as indicators for the work and money domains in particular (to test the Marxist feminist argument more rigorously). Mirroring the women’s absolute achievement indicator, we created the men’s absolute achievement indicator, which allows us to test the male privilege protection hypothesis. These indexes have values ranging from 0 to 1, with 0 indicating no achievement and 1 indicating full achievement. For further details, see Annexes 1 and 2. Independent variables at the individual level Regarding the independent variables, we include variables that were introduced in previous studies using the FRA’s EU-VAW survey (Humbert et al. 2021; Ivert et al. 2019; Reichel, 2017), such as (1) socio-demographic characteristics of women, (2) other characteristics of women (of which some of we improve), (3) couple characteristics, and (4) men’s characteristics. First, we 8 include women’s age, number of children, and employment status. Although these three variables turned out to be insignificant in our models, we have kept them in order to control for possible composition effects in the age, family, and employment structures of countries. Second, we also include other characteristics of women that previous research has demonstrated to have an impact on IPVAW: women’s education, child abuse, marital status, residential area type, and subjective perception of the household’s income situation (Ivert et al. 2019; Humbert et al. 2021; Till- Tentschert, 2017). We improved the measurement of some of these variables: (a) we look at the frequency of child abuse rather than the experience of it (yes/no); and (b) we extend the marital status variable to create a three-category variable: married or cohabiting; having a romantic partner but not living with him; and separated. In addition to these characteristics of previous research, we consider whether the respondent feels she belongs to a minority group. This is a summary variable which indicates that women declare that they consider that they belong to at least one of the following minority groups: ethnic, immigrant, religious, sexual, in terms of disability, or any other minority.2 Third, we also look at three characteristics of the couple: the duration of the relationship, as repetition would be expected to increase in longer relationships; relative earnings, which reflect status inconsistency at the individual level; and an indication of each partner’s power with respect to the couple’s consumption patterns, which reflects the couple’s egalitarian attitudes towards decision-making. Fourth and finally, since the focus of this paper is also on the role of men, we take into account the male partner’s educational level and employment status, as well as the frequency in which he gets drunk. Previous studies have indicated that being often drunk has a role in triggering violence (not on who commits, but when). Even though the evidence on to what extent such an effect would be causal is mixed, studies have shown that alcohol influences the circumstances under which men are likely to commit IPVAW. Men who committed IPVAW when being drunk more often justified their aggressive and violent behaviour, as well as their belief surrounding the role of alcohol related to sexual behaviour (Abbey, 2011; Reichel, 2017; Humbert et al., 2021). [[Insert Table 1 about here]] 2 There is somewhat overlap between the different ‘types’ of minorities, especially with the last item (belonging to any other minority). Actually, it turns out that all of the different categories have been coded as ‘belonging to any other minority’. In addition, there are 1104 respondents who did not consider themselves to be one of the mentioned groups, but they mentioned to belong to another minority. If we would have separated all categories, we would have in many countries only 1 or 2 cases, so we decided in the end to group them all together. 9 Hypothesis testing Derived from these theoretical approaches, we formulate the following hypotheses: Hypothesis 1. A positive effect of gender equality on repetitive IPVAW, which lends support to the ameliorative hypothesis as proposed by the liberal feminist scholars. The indicator which more accurately accounts for gender equality is the Gender Gap Measure, thus a positive effect of this in repetitive IPVAW would lend support to this hypothesis. Hypothesis 2. A negative effect of the different measures of gender equality on repetitive IPVAW, which provides support to the backlash hypothesis as proposed by the radical feminist scholars. A negative effect of the Gender Gap Measure on repetitive IPVAW would provide support to this hypothesis. Hypothesis 3. A negative effect of women’s absolute financial and labour status on repetitive IPVAW, which provides support to the Marxist feminist theorists. Hypothesis 4. A negative effect of men’s absolute financial and labour status on economic achievement on repetitive IPVAW, which provides support to the male privilege protection hypothesis. Analytical strategy We perform multivariate analyses using stepwise multilevel linear regression models. The operationalisation and descriptives of all the variables are provided in Table 1. Our final analyses cover eight models, seven of which we present in Table 2 (Model 0 has only individual level variables and their coefficients are similar to models 1 to 7). Based on these theoretical and methodological considerations, we take the following approach: First, we test the impact of the EIGE’s GEI index on IPVAW (Model 1). Second, we decompose the GEI into two conceptually distinguishable components: one operationalising the Gender Gap Measure (Model 2) and the other one operationalising Overall Population Attainment Measure (Model 3). Introducing the first component into the analysis leads us to rigorously assess the ameliorative effects versus backlash debate. In turn, the Overall Population Attainment component can be further decomposed into the two components of women’s attainment (Model 4) which allows us to test predictions from the Marxist feminist theories, and men’s achievement which allow us to test de male privilege protection hypothesis (Model 5). Because these Marxist feminist theories mainly focus on women’s financial status (Model 6), we additionally narrow down the achievement component to the ‘work’ and ‘money’ domains. Finally, we focus on the role of men’s status (Model 7). 10 [[Insert Table 2 about here]] Results Table 2 presents the models mentioned above. The results at the individual level are stable across all models. The individual-level results indicate that a woman was more likely to endure repetitive acts of violence from her current partner if she had experienced more repetitive violence during her childhood, was lower educated, belonged to a minority group, or was living in a household where she was finding it difficult to cope with the present income. In terms of the partner’s characteristics, a woman was more likely to experience repetitive violence if her partner was lower educated and was getting drunk frequently. Having had shorter and non-cohabiting previous relationships tended to protect a woman from violence, especially if those relationships had been more egalitarian in terms of decisions about how to spend income. On the contextual level, there is little variance to be explained, as only 0.9% of the total individual and macro-level variance takes place at the context level (there is no difference in the intraclass correlation coefficient of models depending on whether it does or does not include composition effects; i.e., whether individual-level variables are included in the ‘empty’ model). All of our macro-indicators of interest explain a little bit of this contextual-level variation, with men’s absolute overall status for the six domains explaining the most (33% of the total variance at the macro level). The first model introducing contextual variables (M1) shows a negative and significant relationship between the EIGE’s Gender Equality Index - GEI (merging gaps and overall achievement) and violence repetition (p-value<0.01). We observe a small effect (b=-2.273); a one standard deviation change in the original GEI index goes together with a (0.087*-2.273) = -0.20 point difference on the repetitive partner violence scale (which has values ranging from 0 to 86, and has a standard deviation of 4.4). The maximum effect that the original GEI index can explain is [(0.80-0.50)*-2.273]3 = -0.68 points on the IPVAW repetition-sensitive indicator, which is still very little. Despite the small effect size, the effect is in the opposite direction of the backlash hypothesis, which predicted that more equality would go together with more IPVAW. However, the next model (M2) shows that the Gender Gap Measure has no significant impact on violence repetition (b=-2.153). In contrast, the Overall Population Attainment Measure (M3) has a negative and statistically significant impact (b=-1.662; p-value<0.001). These results indicate that the Overall Population Attainment component is leading the EIGE’s GEI influence on IPVAW. 3 The EIGE’s GEI has values ranging from 50 to 80, see Table 1. 11 Finally, we find that men’s absolute overall status (M5) shows a stronger effect (b=-2.352; p- value<0.001) than women’s absolute overall status (M4; b=-1.481; p-value<0.01), which points to the importance of the men’s status in IPVAW. In contrast, when we assess men’s and women’s work and money composite indicator, we notice the opposite pattern, with M6 indicating that the impact of the women’s effect (-2.138; p<0.01) is larger than the men’s effect (-1.918; p<0.01), as shown in M7. Robustness tests We performed a couple of robustness checks with respect to the effect sizes of our macro-level indicators (see results in Annex Figure 3). We examine both methodological and theoretical issues. More specifically, we consider: the importance of the method of first contact to conduct the interview; the response rate; and overall violence in a country measured using homicide rates. We also did some outlier analyses. Below we discuss each of these corrections in detail. First, with respect to the methodological issues of the FRA survey, previous studies highlighted that some limitations of the survey make it difficult to draw reliable conclusions from its results. Sylvia Walby and colleagues (Walby et al., 2017; Walby & Towers, 2017) pointed out that the method of first contact constitutes one of the main limitations. While women were first approached by phone in the Nordic countries, women were first approached face to face in most of the EU countries. Although the whole interview was not conducted by phone in the Nordic countries, women had the opportunity to accept or refuse to participate in the survey, which would have affected the (selective) response rates; i.e., women who were less motivated to talk about their experiences were more likely to drop out when given the opportunity to say ‘no’ on the phone. We take this aspect into account by controlling for the method of first contact, and also include those countries where women were first approached by letter (Malta, Slovenia, and the United Kingdom) (FRA, 2014: 13), which is not usually considered in the debate. Variations in the response rate have also been identified as a methodological drawback of the FRA’s EU-VAW survey (Walby & Towers, 2017; Walby et al; 2017). Although the response rate was related to the method of first contact, we also considered this point in the analysis. We found that the method of first contact for conducting the survey interview hardly affected the coefficients of the EIGE’s GEI, the Gender Gap Measure, or women’s or men’s status. Moreover, the country response rate did not appear to change our conclusions. Indeed, if anything, these methodological ‘correction’ variables only increased the negative effect sizes of the attainment variables (see the coefficients for Models “meth”). 12 With respect to the theoretical control, we note that, following previous studies, it could also be argued that violence against women is higher in those countries where violence is more normalised (Humbert et al., 2021). We therefore take into account the most reliable indicator of overall country-wide violent victimisation: homicide rates (UNODC, 2021). While other types of violent crimes, such as non-lethal violent crimes or robbery, are widely underreported, lethal violence is a robust indicator (Krug et al., 2002; UNODC, 2013). Again, we did not find any significant impact of this indicator on IPVAW, or a substantial change in the impact of the overall population attainment indicator or one of its subindices on IPVAW. Third and finally, we checked to what extent certain outliers might affect our results. When looking at the extent IPVAW repetition, we see that Romania, which is a country with a large gender gap and low achievement, has an extremely high score on the IPVAW indicator. We therefore ran models excluding this country, but found no substantive changes in our models. This also applies to the three Nordic countries included in the sample (Denmark, Finland, and Sweden). When excluding these three countries –which are the only ones for which the telephone instead of a letter or a face-to-face visit was used as a first contact method– we did not find substantive changes in the coefficients of our variables of interest. Results are available upon request. Discussion Previous research analysing the relationship between gender equality and violent victimisation using FRA’s EU-VAW survey data and the EIGE’s Gender Equality Index-GEI (Ivert et al., 2019; Sanz-Barbero et al., 2018; Humbert et al., 2021) reported either inconclusive findings or a positive relationship, and thus supported the backlash hypothesis. In this paper, we not only tested the impact of the EIGE’s GEI index on an improved IPVAW repetition-sensitive indicator; we also improved the measurement of gender inequality, thereby distinguishing two components of the EIGE’s GEI: (1) a Gender Gap Measure and (2) an achievement component, using the Overall Population Attainment Measure. In addition, this latter component was included in our models as an overall population attainment measure, and separately for men and women. In this way, we could test the Marxist feminist and the male privilege protection hypothesis. Our results highlight the following points. First, in contrast to previous results confirming the backlash hypothesis, we found the opposite effect: namely, a negative and statistically significant relationship between the EIGE’s GEI and the repetition-sensitive indicator of current partner IPVAW. Therefore, if EIGE’s GEI is considered as a proper measure of gender inequality, our results would support the ameliorative hypothesis. Thus, we can conclude that support for either the backlash or the ameliorative hypothesis is very sensitive to the IPVAW indicator used. This 13 finding on the impact of different indicators on drawing different conclusions is aligned with previous studies (Permanyer, & Gómez-Casillas, 2020; Walby, Towers & Francis, 2016). Second, when we separated the EIGE’s GEI into two different composite indicators, we found that the Gender Gap Measure did not have a significant coefficient, and thus provided no support for either the ameliorative or the backlash hypothesis. Therefore, the results on the relationship between gender equality and IPVAW repetition are also sensitive to how gender equality is measured. On the contrary, we found that the Overall Population Achievement was the component that most explained the lower rates of IPVAW repetition. This finding could be connected with the fact that in countries more committed to investing in social policies, namely the social democratic ones, violence is lower than in liberal ones (Walby, 2009). Hence, higher social expenditure and welfare in general would be a critical aspect leading to overall population achievement, thus impacting in lower victimization rates. Third, our results on the negative and statistically significant effect of women’s absolute status on IPVAW repetition support the Marxist feminist position. These results contrast with Stamatel’s (2018), who didn’t find any significant results for women’s absolute status in the European Union, but comparisons with this study are challenging due to the differences in methodologies and indicators used. Our results could be understood following Dugan et al. (2003), who state that beyond a woman’s specific status and circumstances, her perception of her potential access to resources may discourage her from or encourage her to leave a threatening relationship. In other words, mirroring contributions in the field of union dissolution (van Damme and Kalmijn, 2014), it could be argued that it is not only the actual economic costs of leaving an abusive relationship that matter, but also the expected economic costs of leaving. Thus, women who live in contexts where women have achieved higher levels of economic independence, even if they are not employed or have low occupational status or low earnings, may be expected to leave an abusive relationship, as their expected economic costs of leaving are low in such contexts. Fourth, regarding men’s absolute status, our results lend support to the male privilege protection hypothesis. Although we found that women’s absolute economic status was more important than men’s, for overall achievement for the six domains (‘work’, ‘money’, ‘knowledge’, ‘time’, ‘power’, and ‘health’), we found that men’s achievement has a larger effect than women’s. These apparently mixed results supporting both the Marxist feminist position and male privilege protection hypothesis may be coherently explained considering that in the European Union, women structural economic and labour status is what allows them to leave the violent relationship and, at the same time, women are ‘allowed’ to gain in these domains if the overall attainment of males in considering all the domains (‘work’, ‘money’, ‘knowledge’, ‘time’, ‘power’, and ‘health’) is not threatened. These results could indicate that men, as a gender class, are not equally prepared for gender equality in all domains. Here we should distinguish between a structural 14 position derived from economic and labour status and that derived from overall status (also including the domains ‘knowledge’, ‘time’, ‘power’, and ‘health’). Note that we refer here to men’s overall achievement levels at a country level, and individual men in those countries (contexts) would be affected by the overall levels of achievement of males in that context (contextual effect) –even though there might also be a (small) composition effect left of men’s status that we do not capture by controlling for men’s education and men’s employment status.4 Our study relies on cross-sectional data, which prevents us from providing conclusions over time. With this disclaimer in mind, we can allow some theoretical considerations to be tested in future studies using longitudinal data. The male privilege protection hypothesis extends previous contributions considering that gains in gender equality will go through phases of backlash (Brownmiller, 1975; Russell, 1975) and articulates with the Marxist feminist hypothesis. Both would theoretically explain why there are periods of gains in women’s rights in certain areas in an amelioration process, and how they interlock with periods of backlash. Hence, as women’s gains in terms of equality impact on increased levels of gender equality, a backlash to these newly acquired benefits for women might occur if men feel that they are losing too much of their absolute overall status. For men, it would be no reaction in terms of IPVAW if women’s status improves in specific areas, such as economic and work-related domains, as long as they do not perceive that their own position does not decrease in absolute terms. In sum, following Whaley et al. (2013) and Stamatel (2018), who already acknowledged that IPVAW arises from an interplay of counterbalancing forces that provoke either ameliorative or backlash effects, we provide an explanation that could account for both processes, which also entails understanding gains in absolute status for both men and women and how this would allow an amelioration process or triggering backlash effects. Similarly, this status-related process could explain the increase in violence against women repetition during the 2008 financial crisis in England and Wales (Walby, Towers & Francis, 2016). This study has several limitations, which lead us to be cautious in our conclusions. These limitations have to do with both methodological and theoretical aspects. Methodologically, the most important limitations are related to the drawbacks of the FRA’s EU-VAW survey, such as problems with how the respondents were approached; the questionnaires and the scales used to gather information on victimisation; the sample sizes and sample frames; and the capping of the number of violent incidents included (Walby et al., 2017; Walby and Towers, 2017, Permanyer 4 Macro-level effects can occur due to both contextual level and individual composition effects. In this paper, we do not aim to distinguish between the two, as this would require more data on individual level socio-economic, time, and well-being characteristics of men and women (i.e. for each of the six EIGE GEI dimensions individual level characteristics). However, our reasoning mainly refers to contextual level effects. We leave disentangling composition from contextual effects to future research. 15 & Gómez-Casillas, 2020; Humbert et al., 2021). Due to the small sample size, we could not analyse different sources of minority separately, and we had to collapse them into one variable. Theoretically, we proposed several hypotheses about processes over time that should ideally be tested using cross-national panel data. In this context, it is worth mentioning that since Yllö’s (1984) study first appeared, there has been much discussion about the backlash and ameliorative effects across geographies, while the theoretical interpretations often relied on gender equality advances over time. In other words, different geographies are often considered to be at different stages of development in the road leading to gender equality. For instance, the Nordic countries are considered in a higher stage of “history” due to their higher levels of gender equality. However, Wemrell’s (2022) study of backlash shows that Sweden should not be portrayed as having reached an ultimate phase of development in terms of gender equality. Thus, our paper has limitations related to the fact that we needed to rely on cross-sectional data. We would therefore like to call for a new data collection effort that not only updates the 2012 FRA data, but also looks at longitudinal processes. Still, our male privilege protection hypothesis could be helpful for understanding the mixed results that appear to support both ameliorative and backlash effects. Additionally, due to data availability, we were not able to account for the cultural-specific issues that shape women’s disclosure of their experiences of violence in a country. Using FRA’s EU-VAW survey question asking respondents if they know any women who has suffered any form of domestic violence to account for women’s disclosure at a country level would not be appropriate because women would report a higher knowledge of women in this situation in the countries where violent partners re-partner more often, as explained by the violent partners’ rotation hypothesis (Gómez-Casillas, 2018; Permanyer, & Gómez-Casillas, 2020). Finally, we focus our study on the European Union Member States, where levels of gender equality are quite high, which calls for being cautious in not extrapolating our conclusions to other contexts. Conclusions Our results do not clearly support either the backlash or the ameliorative hypothesis when gender equality is operationalized using the Gender Gap Measure. They do, however, present two other important findings: First, country-level women’s status, and especially women’s financial and employment-related status, have an effect on lessening IPVAW, thus providing support for the Marxist feminist theoretical positions. In countries where women’s economic and labour status is higher, women would be encouraged to leave a violent relationship not only due to each women’s actual status but also due to her potential access to resources after a separation. Second, our results indicate that men’s overall status has an important impact on lower rates of IPVAW, lending support to the ‘male privilege protection’ hypothesis. Interestingly, the effect of women’s status in the ‘work’ and ‘money’ domains is larger than men’s, while the effect of men’s overall status 16 for the six domains (‘work’, ‘money’, ‘knowledge’, ‘time’, ‘power’, and ‘health’) is higher than women’s. With respect to men, in countries where their overall status is higher, they would feel that their privilege as members of a ‘gender class’ is protected and therefore women's status gains in economic and financial terms would not pose a great threat to them. This suggests that men’s acceptance of women’s status acquisition is not linear and not uniform across sources of status, which would involve an interplay of backlash and ameliorative process. Acknowledgments, credits and funding: The study was supported by the “Pacto de Estado contra la Violencia de género”, Ministerio de la Presidencia, Relaciones con las Cortes e Igualdad, Secretaría de Estado de Igualdad. Iñaki Permanyer acknowledges financial support from the European Research Council (Grant No. 864616). References Abbey, A. (2011). Alcohol's role in sexual violence perpetration: Theoretical explanations, existing evidence and future directions. Drug and alcohol review, 30(5), 481-489. Brownmiller, S. (1975). Against Our Will: Men, Women and Rape. New York: Simon & Schuster. Connell, R. W., & Messerschmidt, J. W. (2005). Hegemonic masculinity: Rethinking the concept. Gender & society, 19(6), 829-859. Daly, M. (1978). Gyn/Ecology: The Metaethics of Radical Feminism. London: Women’s Press. Demetriou, D. Z. (2001). Connell's concept of hegemonic masculinity: A critique. Theory and society, 30(3), 337-361. Dobash R., & Dobash, E. (1979). Violence Against Wives: A Case Against Patriarchy. New York: Free Press. Donaldson, M. (1993). What is hegemonic masculinity?. Theory and society, 22(5), 643-657. 17 Dugan, L., Nagin, D. S., & Rosenfeld, R. (2003). Exposure reduction or retaliation? The effects of domestic violence resources on intimate‐partner homicide. Law & society review, 37(1), 169-198. EIGE—European Institute for Gender Equality. (2021a). Gender Equality Index scores, domain scores and sub-domain scores for 2012 [data collection] Retrieved from https://eige.europa.eu/gender-statistics/dgs/indicator/index_data__index_scores EIGE—European Institute for Gender Equality. (2017). Gender Equality Index 2017. Methodological Report. Luxembourg: Publications Office of the European Union. Esses, V. M., Dovidio, J. F., Jackson, L. M., & Armstrong, T. L. (2001). The immigration dilemma: The role of perceived group competition, ethnic prejudice, and national identity. Journal of Social issues, 57(3), 389-412. FRA—European Union Agency for Fundamental Rights. (2015). European Union Agency for Fundamental Rights (FRA): Violence Against Women survey, 2012: Special Licence Access. [data collection]. UK Data Service. SN: 7730. http://doi.org/10.5255/UKDA-SN- 7730-1 FRA—European Union Agency for Fundamental Rights. (2014). Violence against women: An EU-wide survey. Main results. Luxembourg: Publications Office of the European Union. Gómez-Casillas, A. (2018). Violencia de género y prácticas tradicionales perjudiciales: una contribución al debate metodológico para su medición en Europa. Universitat Autonoma de Barcelona. Cerdanyola del Valles [Thesis] Gracia, E., & Merlo, J. (2016). Intimate partner violence against women and the Nordic paradox. Social Science & Medicine, 157, 27-30. Gracia, E., Martín-Fernández M., Lila, M., Merlo, J., & Ivert, A.K. (2019). Prevalence of intimate partner violence against women in Sweden and Spain: a psychometric study of the ‘‘Nordic paradox’’. PLoS ONE 14(5),1–17. Humbert, A. L., Strid, S., Hearn, J., Balkmar, D. (2021). Undoing the ‘Nordic Paradox’: Factors affecting rates of disclosed violence against women across the EU. PLoS ONE, 16(5), e0249693. Ivert, A. K., Gracia, E., Lila, M., Wemrell, M., & Merlo, J. (2019). Does country-level gender equality explain individual risk of intimate partner violence against women? A multilevel analysis of individual heterogeneity and discriminatory accuracy (MAIHDA) in the European Union. European journal of public health, 30(2), 293-299. Jaggar, A. M. (1983). Feminist politics and human nature. Totowa, NJ: Rowman & Littlefield. 18 Kalmuss, D. S., & Straus, M. A. (1982). Wife's marital dependency and wife abuse. Journal of Marriage and the Family, 277-286. Kosakowska‐Berezecka, N., Besta, T., Bosson, J. K., Jurek, P., Vandello, J. A., Best, D. L., et al. (2020). Country‐level and individual‐level predictors of men's support for gender equality in 42 countries. European Journal of Social Psychology, 50(6), 1276-1291. Krug, E. G., Dahlberg, L. L., Mercy, J. A., Zwi, A. B., & Lozano, R. (2002). World report on violence and health. Geneve: World Health Organization. Permanyer, I., & Gomez-Casillas, A. (2020). Is the ‘Nordic Paradox’ an illusion? Measuring intimate partner violence against women in Europe. International journal of public health, 65, 1169–1179. Permanyer, I. (2015). Why call it ‘equality’ when it should be ‘achievement’? A proposal to un- correct the ‘corrected gender gaps’ in the EU Gender Equality Index. Journal of European Social Policy, 25(4), 414-430. Reichel, D. (2017). Determinants of intimate partner violence in Europe: The role of socioeconomic status, inequality, and partner behavior. Journal of interpersonal violence, 32(12), 1853-1873. Russell, D. (1975). The Politics of Rape, the Victim’s Perspective. New York: Stein & Day. Sanz-Barbero, B., Corradi, C., Otero-García, L., Ayala, A., & Vives-Cases, C. (2018). The effect of macrosocial policies on violence against women: a multilevel study in 28 European countries. International Journal of Public Health, 63(8), 901-911. Stamatel, J. P. (2018). Money matters: Dissecting the relationship between gender equality and female homicide victimization rates in the European Union. Feminist Criminology, 13(5), 435-455. Stöckl, H., Devries, K., Rotstein, A., Abrahams, N., Campbell, J., Watts, C., & Moreno, C. G. (2013). The global prevalence of intimate partner homicide: a systematic review. The Lancet, 382(9895), 859-865. Till-Tentschert, U. (2017). The relation between violence experienced in childhood and women’s exposure to violence in later life: Evidence from Europe. Journal of interpersonal violence, 32(12), 1874-1894. UNODC – United Nations Office on Drugs and Crime. (2021). Intentional homicide and sexual offences by legal status and sex of the person involved - number and rate for the relevant sex group. [data collection] Retrieved from https://dataunodc.un.org/content/data/homicide/homicide-rate 19 UNODC – United Nations Office on Drugs and Crime. (2013). Global Study on Homicide 2013. Trends, contexts, data. Viena: UNODC. van Damme, M., & Kalmijn, M. (2014). The dynamic relationships between union dissolution and women’s employment: A life-history analysis of 16 countries. Social Science Research, 48(0), 261-278. Vieraitis, L. M., Britto, S., & Morris, R. G. (2015). Assessing the impact of changes in gender equality on female homicide victimization: 1980-2000. Crime & Delinquency, 61(3), 428- 453. Walby, S., Towers, J., Balderston, S., Corradi, C., Francis, B., Heiskanen, M., et al. (2017). The concept and measurement of violence against women and men. Bristol: Policy Press. Walby, S, & Towers, J. (2017). Measuring violence to end violence: mainstreaming gender. Journal of Gender-Based Violence, 1(1), 11–31. Walby, S., Towers, J., & Francis, B. (2016). Is violent crime increasing or decreasing? A new methodology to measure repeat attacks making visible the significance of gender and domestic relations. British Journal of Criminology, 56(6), 1203-1234. Walby, S. (2009). Globalization and inequalities: Complexity and contested modernities. London: Sage. Wemrell, M. (2022). Stories of backlash in interviews with survivors of intimate partner violence against women in Sweden. Violence against women, doi: 10778012221088312. Whaley, R. B., Messner, S. F., & Veysey, B. M. (2013). The relationship between gender equality and rates of inter-and intra-sexual lethal violence: An exploration of functional form. Justice Quarterly, 30(4), 732-754. Yllö, K. (1984). The status of women, marital equality, and violence against wives: A contextual analysis. Journal of family Issues, 5(3), 307-320. 20 Table 1 Table 1. Definitions of measures used in the analyses and descriptive statistics, FRA (mean and standard deviation / percentages). Analytical sample – missing cases listwise deleted: N = 27,195 women. Variable Definition Descriptive statistics Percentage or mean Mini- Maxi- (standard deviation) mum mum Main analyses IPV repeated Continuous variable, normalized sum score of answering categories 0.77 (4.4) 0 86 current regarding its repetition (0, 1, 3.5, 6) on each of the 13 items: ‘pushed partner you or shoved you’; ‘slapped you; threw a hard object at you’; ‘grabbed you or pulled your hair’; ‘beat you with a fist or a hard object, or kicked you’; ‘burned you’; ‘tried to suffocate you or strangle you’; ‘cut or stabbed you, or shot at you’; ‘beat your head against something’; ‘being forced you into sexual intercourse’; ‘attempt of forced sexual intercourse’; ‘made you take part in any form of sexual activity when you did not want to, or you were unable to refuse’; ‘consent to sexual activity because you were afraid of what might happen if you refused’ Age of Asked in seven categories: 1. 18-24; 2. 25-29; 3. 30-34; 4. 35-39; 5. 40- 18-24: 6.4; 1 7 respondent 49; 6.50-59; 7. 60+ 25-29: 7.7 30-34: 9.7; 35-39: 11; 4049: 23; 50-59: 21; 60+: 22 Childhood Frequency of physical and or sexual violence during childhood 0.048 (0.10) 0 1 abuse (continuous variable) based on the following question: “The next questions concern the time when you were under 15 years of age. Before you were 15 years old, how often did any adult – that is, somebody who was 18 years or older – do any of the following to you: Slap you or pull your hair so that it hurt you? Hit you very hard so that it hurt you? Kick you very hard so that it hurt you? Beat you very hard with an object like a stick, cane or belt so that it hurt you? Stab or cut you with something so that it hurt you? Expose their genitals to you when you did not want them to? Make you pose naked in front of any person or in photographs, video or an internet webcam? Touch your private parts - genitals or breasts - when you did not want them to? Make you touch their private parts - genitals or breasts - when you did not want to? Have sexual intercourse with them when you did not want to? Answering categories: Never (0), once (1), more than once (3). Continuous variable, normalized sum score of answering categories regarding its repetition (0,1,3) on each of the 10 items. Education Highest ISCED level attained: 0 “pre-primary education”, 1 “primary high: 23; 1 3 respondent education”, 2 “lower secondary education”, 3 “(upper) secondary medium: 50; education”, 4 “post-secondary non tertiary education”, 5 “first stage of low: 27 tertiary education (not leading directly to an advanced research qualification)”, 6 “second stage of tertiary education (leading to an advanced research qualification)”. Recoded into 1 "low: level 0-2" 2 "mid: level 3-4" 3 "high: level 5-6". Education Idem Education respondent high: 22; 1 3 partner medium: 44; low: 35 Employment Recoded from the following question: “Are you working in a paid job or full-time: 42; 1 7 status are you unemployed or doing something else – using this card, how part-time: 12; would you describe your main activity?” Final categories: 1. “Paid homemaker: 16; work: Full-time or self-employed”, 2. “Paid work – part-time”, 3 unemployed: 8.2; "Homemaker – looking after the home/children/relatives etc. or on student: 3.7; parental leave, other unpaid or voluntary work", 4. "Unemployed”, 5 disabled: 1.2; "Student – in training", 6 “Disabled”, 7 "Retired", . Missing (military, retired: 16 other) Employment Idem Employment status respondent full-time: 67; 1 7 status partner part-time: 2.5; homemaker: 0.4; unemployed: 4.7; student: 2.0; 21 disabled: 1.4; retired: 22 Living area 1 “Town or village”, 2 “Farm”, 3 “City”. Town: 63; 1 3 Farm: 3.3; City: 34 Children 0 “No children”, 1 “1 Child”, 2 “2 Children”, 3 “3+ Children” 0: 54; 0 3 1:21; 2: 18’ 3+: 7.0 Minority Categorical variable; checking at least one or more of the categories (1-5 None: 88; 0 2 or 7) based on the following question: Thinking about where you live, 1: 12 do you consider yourself to be part of any of the following? Please tell me all that apply: 1. An ethnic minority; 2. An immigrant minority; 3. A religious minority; 4. A sexual minority; 5. A minority in terms of disability; 6. Any other minority group, specify …; 7 None. Recoded: 0. Not belonging to a minority group, 1. Belonging to at least one minority group. Relative Would you say that your partner earns less than you, or your earnings he earns more: 63; 1 4 earnings are roughly the same, or that your partner earns more than you? 1 “He equal: 21; earns more”, 2 “Equal earning”, 3 “She earns more”, 4 “missing (dk, she earns more: 12; nap, refused)” missing: 4.5 Income Which of the descriptions on this card comes closest to how you feel living comfortably: 1 3 coping about your household’s income nowadays? 1 “Living comfortably on 25; income”, 2”Coping on income”, 3 “Finding it difficult or very difficult Coping: 43; on income Difficult: 32 Relationship 1 “Married or civil partnership; 2 “romantic relationship, not living Married: 91; 1 3 type together”, 3 “separated” Relationship: 9.1; Separated: 0.3 Equal say Do you feel you have an equal say with regard to the use of the No equal say: 6.5; 1 3 household income? 1 “Yes”, 2 “No”, 3 “missing (don’t know, not Equal say: 83; applicable, refused)” missing: 11 Union Asked in seven categories: 1. <1 year; 2. 1-10; 3. 11-20; 4. 21-30; 5. 31- <1: 2.3; 1 7 duration 40; 6.41-50; 7. 51+ 1-10: 28; 11-20: 21; 21-30: 18; 31-40: 16; 41-50: 11; 51+: 2.6 Partner drunk Treated as continuous variable: How often does your partner drink so 1. Never: 60; 1 4 much that he/she gets drunk? 1 “Never drinks”, 2 “A few times a year or 2. Up to once a less”, 3 “Once every two months”, 4 “Once a month” 5 “A couple of month: 32; times a month”, 6 “Once or twice a week”, 7 “Every day or almost every 3. Up to twice a day”. Recoded into: 1. Never, 2. Up to once a month; 3. Up to twice a week: 6.4; week, 4. Every day. 4. Every day: 0.9 Gender This index (2012), created by EIGE, is a composite measure of six 61 (8.6) 50 80 Equality domains (work, money, knowledge, time, power, health). Together it Index consists of 31 indicators (when it was first released it measured 27 (EIGE’s GEI) indicators) which measure both the Gender Gap and Women’s Achievement. . It originally ranges from 1 to 100. Source: Database sent by EIGE under request Gender Gap This index (2012) is the Gender Gap component of EIGE’s GEI, 72 (6.0) 60 84 Measure uncontrolled for Achievement in the six domains. It ranges from 0 to 1, meaning 0 full inequality and 1 full equality. Source: Permanyer (2015). Overall This index (2012) is the Overall Population Attainment component of 64 (14) 38 90 Population EIGE’s GEI in the six domains. It ranges from 0 to 1, meaning 0 no Attainment achievement and 1 full achievement. Measure Women’s This index (2012) is the Overall Women’s achievement in the six 56 (14) 35 88 absolute domains. It ranges from 0 to 1, meaning 0 full inequality and 1 full achievement equality. Men’s This index (2012) is the Overall Men’s achievement in the six domains. 67 (10) 45 85 absolute It ranges from 0 to 1, meaning 0 full inequality and 1 full equality. achievement Women’s abs. This index (2012) is the Overall Women’s achievement in the two of the 72 (9.5) 57 88 achievement six EIGE’s GEI domains: ‘work’ and ‘money’. It ranges from 0 to 1, work and meaning 0 full inequality and 1 full equality. 22 money domain Men’s abs. This index (2012) is the Overall Men’s achievement in the two of the six 74 (11) 58 91 achievement EIGE’s GEI domains: ‘work’ and ‘money’. It ranges from 0 to 1, work and meaning 0 full inequality and 1 full equality. money domain Supplementary analyses (Annex 3) Contact Based on table 4.2 of FRA’s EU VAW Survey technical annex: method Face-to-face: 77 1 3 method used for contacting the sampled addresses and individuals for the first Telephone: 11 time, by EU Member State. 1. Directly face-to-face; 2. By telephone; 3. Letter: 11 By letter Response Based on table 7.2 of FRA’s EU VAW Survey technical annex: method 48 (16) 19 84 Rate used for contacting the sampled addresses and individuals for the first time, by EU Member State. Homicide Owns calculations for the year 2011 (average 2010-2012) based on 1.52 (1.20) 0.73 6.8 victims of intentional homicide. Homicide rate average 2011-2012. Source: Authors’ elaboration from UNODC, 2021. Source: Authors’ calculations based on the FRA’s survey on Violence Against Women Survey dataset, 2012 and other sources. 23 Table 2 Table 2. Experience of Repeated IPVAW for women who are currently partnered. Multilevel Linear Regression on continuous Repeated IPVAW experience variable (Minimum 0 No IPVAW, maximum 100 Many repeated IVP experiences) (1) (2) (3) (4) (5) (6) (7) Age 18-24 ref ref ref ref ref ref ref Age 25-29 0.154 0.151 0.154 0.153 0.155 0.155 0.155 Age 30-34 0.145 0.143 0.144 0.144 0.145 0.145 0.146 Age 35-39 0.090 0.087 0.090 0.089 0.090 0.089 0.090 Age 40-49 0.169 0.164 0.170 0.168 0.170 0.168 0.169 Age 50-59 0.154 0.148 0.155 0.154 0.154 0.152 0.152 Age 60+ 0.122 0.114 0.126 0.125 0.125 0.120 0.120 Frequency Child Abuse 3.490*** 3.483*** 3.494*** 3.491*** 3.495*** 3.491*** 3.493*** High Educat. Level ref ref ref ref ref ref ref Medium Educat. Level 0.098 0.099 0.095 0.096 0.096 0.098 0.099 Low Educat. Level 0.254** 0.255** 0.250** 0.250** 0.252** 0.255** 0.257** High Educat. Level Partner ref ref ref ref ref ref ref Medium Educat. Level Partner 0.026 0.028 0.025 0.026 0.023 0.025 0.025 Low Educat. Level Partner 0.330*** 0.333*** 0.326*** 0.329*** 0.325*** 0.330*** 0.330*** Full-time or self-employed ref ref ref ref ref ref ref Parttime 0.109 0.100 0.114 0.108 0.119 0.113 0.114 Homemaker/unpaid work -0.030 -0.033 -0.027 -0.031 -0.024 -0.026 -0.024 Unemployed 0.149 0.149 0.149 0.149 0.149 0.148 0.148 Student -0.019 -0.027 -0.015 -0.018 -0.015 -0.020 -0.020 Disabled 0.249 0.241 0.255 0.250 0.257 0.248 0.247 Retired -0.099 -0.098 -0.101 -0.099 -0.102 -0.101 -0.102 Ftime or self-employed Partner ref ref ref ref ref ref ref Parttime Partner 0.232 0.230 0.236 0.233 0.237 0.233 0.233 Homemaker Partner 2.230*** 2.227*** 2.228*** 2.229*** 2.225*** 2.226*** 2.226*** Unemployed Partner 0.443*** 0.441*** 0.444*** 0.443*** 0.443*** 0.442*** 0.441*** Student Partner 0.202 0.198 0.202 0.201 0.203 0.202 0.203 Disabled Partner 0.743*** 0.739*** 0.748*** 0.745*** 0.750*** 0.743*** 0.742*** Retired Partner 0.069 0.070 0.069 0.069 0.069 0.070 0.070 Town or Village ref ref ref ref ref ref ref Farm 0.359* 0.357* 0.361* 0.362* 0.357* 0.354* 0.351* City -0.023 -0.021 -0.025 -0.023 -0.027 -0.025 -0.025 No Children ref ref ref ref ref ref ref 1 Child 0.025 0.025 0.024 0.025 0.024 0.025 0.024 2 Children 0.045 0.044 0.044 0.045 0.044 0.044 0.044 3+ Children 0.121 0.118 0.122 0.121 0.122 0.120 0.119 Not belonging to a minority ref ref ref ref ref ref ref Belonging to at least one 0.247** 0.245** 0.246** 0.245** 0.247** 0.247** 0.247** minority He earns more ref ref ref ref ref ref ref Equal Earning 0.050 0.051 0.050 0.050 0.050 0.050 0.050 She earns more 0.015 0.013 0.017 0.016 0.017 0.015 0.015 Earnings missing 0.139 0.142 0.138 0.138 0.138 0.139 0.139 Living comfortably on present ref ref ref ref ref ref ref income Coping on present income -0.058 -0.049 -0.063 -0.058 -0.065 -0.061 -0.061 Finding it difficult or very 0.323*** 0.338*** 0.317*** 0.324*** 0.313*** 0.319*** 0.319*** difficult on present income Married or civil partnership ref ref ref ref ref ref ref Romantic relationship -0.874*** -0.879*** -0.873*** -0.876*** -0.871*** -0.875*** -0.873*** Separated -0.175 -0.178 -0.171 -0.174 -0.169 -0.175 -0.174 No equal say in how to spend ref ref ref ref ref ref ref income Equal say in how to spend -3.027*** -3.030*** -3.025*** -3.026*** -3.025*** -3.028*** -3.028*** income Equal say income Missing -2.179*** -2.177*** -2.178*** -2.177*** -2.179*** -2.176*** -2.177*** Duration union less than 1 year ref ref ref ref ref ref ref Duration relationship 1-10 years 0.026 0.028 0.025 0.026 0.024 0.025 0.026 Duration relationship 11-20 0.293 0.296 0.291 0.292 0.291 0.292 0.293 years Duration relationship 21-30 0.320 0.324 0.319 0.320 0.319 0.321 0.321 years Duration relationship 31-40 0.393 0.398 0.391 0.392 0.391 0.393 0.394 24 years Duration relationship 41-50 0.688** 0.693** 0.686** 0.686** 0.687** 0.689** 0.690** years Duration relationship 51+ years 0.906*** 0.908*** 0.906*** 0.905*** 0.907*** 0.908*** 0.910*** Partner never drunk ref ref ref ref ref ref ref Partner drunk up till once a 0.363*** 0.362*** 0.364*** 0.366*** 0.362*** 0.361*** 0.359*** month Partner drunk up till twice a 2.143*** 2.141*** 2.147*** 2.147*** 2.144*** 2.140*** 2.138*** week Partner drunk every day 8.304*** 8.305*** 8.302*** 8.304*** 8.299*** 8.301*** 8.299*** EIGE’s GEI -2.273** Gender Gap Measure -2.153 Overall Population -1.662*** Attainment Measure Women’s abs. achievement -1.481** Men’s abs. achievement -2.352*** Women’s abs. ach. work and -2.138** money domain Men’s abs. ach. work and -1.918** money domain _cons 3.604*** 3.766*** 3.285*** 3.049*** 3.801*** 3.750*** 3.638*** sd country level 0.34 0.37 0.32 0.34 0.31 0.34 0.34 sd individual level 4.13 4.13 4.13 4.13 4.13 4.13 4.13 icc 0.007*** 0.008*** 0.006*** 0.007*** 0.006*** 0.007*** 0.007*** BIC 154831 154835 154827 154830 154826 154830 154830 N 27195 27195 27195 27195 27195 27195 27195 Notes: 1 * p<0.05, ** p<0.01, *** p<0.001 2 Bold variables are country-level variables and their coefficients Source: Authors’ calculations based on the FRA’s survey on Violence Against Women Survey dataset, 2012 and other sources. 25 Supplemental Annexes: Supplement to: Gómez-Casillas, A.; van Damme, M.; Permanyer, I. Women’s and Men’s Status: Revisiting the Relationship Between Gender Equality and Intimate Partner Violence Against Women in Europe. Journal of Interpersonal Violence https://doi.org/10.1177/08862605231158760 Annex 1 The formula to estimate the Gender Gap Measure, the Overall Population Attainment Measure and the and women’s absolute status is the following EIGE’s aggregation methodology (EIGE, 2017: 17) 1 𝑤𝑑 ϒ 𝑘𝑑 𝐼𝑖𝑡 = ∏𝑛𝑑=1 {∏𝑘𝑠=1 𝑑 [ ∑𝑛𝑣=1 𝑠 ] } (1) 𝑛𝑠 Where 𝐼𝑖𝑡 is the composite indicator for each of the European Union Member States (𝑖) in the specific year (𝑡); 𝑑 is the number of domains, 6 for the Gender Gap Measure, the Overall Population Attainment Measure and the Women’s absolute status and two for the Marxist indicator (work and money domain); 𝑠 represents the subdomains; 𝑛𝑠 the number of indicator in the subdomains s; 𝑘𝑑 is the number subdomain in the domain 𝑑; 𝑤𝑑 is the weight related to the domain as indicated by experts consulted by EIGE. For the Gender Gap Measure: 𝑥𝑤 𝑖𝑡 ϒ(𝑥𝑖𝑡 ) = |𝑥 𝑚 | (2) 𝑖𝑡 𝑤 𝑚 Where 𝑥𝑖𝑡 represent women’s scores while 𝑥𝑖𝑡 represent men’s scores for each of the 28 EU countries (i) in 2012 (t). For the Overall Population Attainment Measure: 𝑎 𝑥𝑖𝑡 ϒ(𝑥𝑖𝑡 ) = (3) 𝑀 𝑎 Where 𝑥𝑖𝑡 represents the overall score, aggregating women and men’s achievement in each of the indicators and 𝑀is the maximum observed value across countries. For the Women’s absolute status index 𝑤 𝑥𝑖𝑡 ϒ(𝑥𝑖𝑡 ) = (4) 𝑀 𝑤 Where 𝑥𝑖𝑡 represent women’s achievement in each for each of the 28 EU countries. 26 Annex 2 Table A1. List of indicators of the Gender Equality Index Relevant for Relevant for Overall women’s and Domain Subdomain N Indicator and reference population Population men’s Attainment absolute indicator status 1 Full-time equivalent employment rate (%, 15+ population) Relevant Relevant Participation 2 Duration of working life (years, 15+ population) Relevant Relevant Employed people in education, human health and social 3 Not relevant Not relevant Work work activities (%, 15+ employed) Segregation and Ability to take an hour or two off during working hours to quality of work 4 Relevant Relevant take care of personal or family matters (%,15+ workers) 5 Career Prospects Index (points, 0-100) Relevant Relevant Financial 6 Mean monthly earnings (PPS, working population) Relevant Relevant resources 7 Mean equivalized net income (PPS, 16+ population) Relevant Relevant Money Economic Not-at-risk-of-poverty, ≥60 % of median income (%,16+ 8 Relevant Relevant situation population) 9 S20/S80 income quintile share (16+ population) Not relevant Not relevant 10 Graduates of tertiary education (%, 15+ population) Relevant Relevant Attainment and participation People participating in formal or non-formal education and 11 Relevant Relevant training (%, 15+ population) Knowledge Tertiary students in the fields of education, health and Segregation 12 welfare, humanities and arts (tertiary students) (%, 15+ Not relevant Not relevant population) People caring for and educating their children or 13 grandchildren, elderly or people with disabilities, every day Not relevant Not relevant Care activities (%, 18+ population) People cooking and/or doing housework, every day (%, 18+ 14 Not relevant Not relevant population) Time Workers doing sporting, cultural or leisure activities outside 15 of their home, at least daily or several times a week (%, 15+ Relevant Relevant Social activities workers) Workers involved in voluntary or charitable activities, at 16 Not relevant Not relevant least once a month (%, 15+ workers) 17 Share of ministers (% W, M) Not available* Relevant Political 18 Share of members of parliament (% W, M) Not available* Relevant 19 Share of members of regional assemblies (% W, M) Not available* Relevant Share of members of boards in largest quoted companies, 20 Not available* Relevant Economic supervisory board or board of directors (% W, M) Power 21 Share of board members of central bank (% W, M) Not available* Relevant Share of board members of research funding organizations 22 Not available* Relevant (% W, M) Share of board members in publicly owned broadcasting Social 23 Not available* Relevant organizations (% W, M) Share of members of highest decision-making body of the 24 Not available* Relevant national Olympic sport organizations (% W, M) Self-perceived health, good or very good (%, 16+ 25 Relevant Relevant population) Status 26 Life expectancy in absolute value at birth (years) Relevant Relevant 27 Healthy life years in absolute value at birth (years) Relevant Relevant People who do not smoke and are not involved in harmful 28 Relevant Relevant drinking (%, 16+ population) Health Behavior People doing physical activities and/or consuming fruits and 29 Relevant Relevant vegetables (%, 16+ population) Population without unmet needs for medical examination 30 Relevant Relevant (%, 16+ population) Access People without unmet needs for dental examination (%, 16+ 31 Relevant Relevant population) Note: * not available for Overall Population Attainment indicator measures Source: Authors’ elaboration from EIGE, 2017 27 Annex 3 Figure A1. Experience of Repeated IPVAW for women who are currently partnered. Multilevel Linear Regression on continuous Repeated IPV experience variable (Minimum 0 No IPV, Maximum 100 Many repeated IPVAW experiences). EIGE GEI Alter GEI Achievement Women achievement Men achievement Women ec achievement Men ec achievement -8 -6 -4 -2 0 2 M1 M2 M3 M4 M5 M6 M7 M1hom M2hom M3hom M4hom M5hom M6hom M7hom M1meth M2meth M3meth M4meth M5meth M6meth M7meth M13var M23var M33var M43var M53var M63var M73var Notes: MXhom=Homicide rate, which operationalise criminal climate; MXmeth= Methodological controls (method of first contact and response rate); MX3var=control all three variables together (More information available in table 1) Source: Authors’ calculations based on the FRA’s survey on Violence Against Women Survey dataset, 2012 and other sources. 28 Annex 4 Table A2. Experience of Repeated IPVAW for women who are currently partnered. Multilevel Linear Regression on continuous Repeated IPV experience variable (Minimum 0 No IPV, maximum 100 Many repeated IPVAW experiences). Sensitivity check using macrolevel data of 2010 instead of 2012. (1) (2) (3) (4) (5) (6) (7) Age 18-24 ref ref ref ref ref ref ref Age 25-29 0.153 0.150 0.155 0.153 0.156 0.155 0.156 Age 30-34 0.144 0.142 0.144 0.144 0.145 0.145 0.147 Age 35-39 0.089 0.086 0.090 0.089 0.090 0.089 0.091 Age 40-49 0.168 0.161 0.170 0.168 0.170 0.168 0.169 Age 50-59 0.153 0.146 0.155 0.154 0.154 0.151 0.152 Age 60+ 0.122 0.112 0.126 0.124 0.124 0.120 0.120 Frequency Child Abuse 3.489*** 3.481*** 3.493*** 3.490*** 3.493*** 3.490*** 3.491*** High Educat. Level ref ref ref ref ref ref ref Medium Educat. Level 0.098 0.100 0.095 0.096 0.097 0.098 0.099 Low Educat. Level 0.252** 0.253** 0.250** 0.250** 0.253** 0.256** 0.259** High Educat. Level Partner ref ref ref ref ref ref ref Medium Educat. Level Partner 0.026 0.028 0.025 0.026 0.024 0.025 0.025 Low Educat. Level Partner 0.331*** 0.334*** 0.326*** 0.329*** 0.325*** 0.330*** 0.330*** Full-time or self-employed ref ref ref ref ref ref ref Parttime 0.107 0.097 0.114 0.107 0.119 0.114 0.115 Homemaker/unpaid work -0.031 -0.033 -0.026 -0.031 -0.023 -0.026 -0.023 Unemployed 0.149 0.149 0.149 0.149 0.149 0.148 0.148 Student -0.021 -0.029 -0.016 -0.019 -0.015 -0.020 -0.020 Disabled 0.249 0.240 0.255 0.250 0.257 0.248 0.246 Retired -0.099 -0.098 -0.101 -0.099 -0.102 -0.101 -0.102 Full-time or self-employed ref ref ref ref ref ref ref Partner Parttime Partner 0.232 0.229 0.236 0.233 0.237 0.234 0.233 Homemaker Partner 2.230*** 2.225*** 2.228*** 2.229*** 2.225*** 2.226*** 2.225*** Unemployed Partner 0.442*** 0.440*** 0.444*** 0.443*** 0.443*** 0.442*** 0.441*** Student Partner 0.201 0.197 0.202 0.201 0.203 0.203 0.203 Disabled Partner 0.743*** 0.738*** 0.748*** 0.744*** 0.750*** 0.742*** 0.741*** Retired Partner 0.069 0.069 0.069 0.068 0.069 0.070 0.070 Town or Village ref ref ref ref ref ref ref Farm 0.358* 0.354* 0.361* 0.361* 0.358* 0.355* 0.351* City -0.022 -0.020 -0.025 -0.022 -0.027 -0.025 -0.025 No Children ref ref ref ref ref ref ref 1 Child 0.025 0.025 0.024 0.025 0.024 0.025 0.024 2 Children 0.045 0.044 0.045 0.045 0.044 0.044 0.044 3+ Children 0.120 0.117 0.122 0.120 0.122 0.120 0.119 Not belonging to a minority ref ref ref ref ref ref ref Belonging to at least one minority 0.246** 0.245** 0.246** 0.245** 0.247** 0.247** 0.247** He earns more ref ref ref ref ref ref ref Equal Earning 0.050 0.051 0.050 0.050 0.050 0.050 0.050 She earns more 0.015 0.012 0.017 0.016 0.017 0.015 0.015 Earnings missing 0.139 0.143 0.138 0.139 0.138 0.139 0.139 Living comfortably on present ref ref ref ref ref ref ref income Coping on present income -0.056 -0.047 -0.063 -0.057 -0.065 -0.061 -0.062 Finding it difficult or very 0.327*** 0.343*** 0.317*** 0.326*** 0.313*** 0.320*** 0.319*** difficult on present income Married or civil partnership ref ref ref ref ref ref ref Romantic relationship -0.875*** -0.881*** -0.873*** -0.876*** -0.871*** -0.875*** -0.872*** Separated -0.176 -0.178 -0.171 -0.175 -0.169 -0.176 -0.175 No equal say in how to spend 0.000 0.000 0.000 0.000 0.000 0.000 0.000 income Equal say in how to spend -3.028*** -3.031*** -3.025*** -3.027*** -3.025*** -3.028*** -3.029*** income Equal say income Missing -2.178*** -2.177*** -2.178*** -2.177*** -2.179*** -2.176*** -2.178*** Duration union less than 1 year ref ref ref ref ref ref ref Duration relationship 1-10 years 0.026 0.029 0.025 0.026 0.024 0.025 0.026 Duration relationship 11-20 years 0.293 0.296 0.291 0.292 0.291 0.292 0.293 Duration relationship 21-30 years 0.320 0.325 0.319 0.320 0.319 0.321 0.322 Duration relationship 31-40 years 0.393 0.399 0.391 0.392 0.391 0.393 0.395 Duration relationship 41-50 years 0.688** 0.693** 0.686** 0.687** 0.687** 0.689** 0.691** Duration relationship 51+ years 0.906*** 0.909*** 0.906*** 0.905*** 0.907*** 0.908*** 0.910*** Partner never drunk ref ref ref ref ref ref ref Partner drunk up till once a 0.364*** 0.362*** 0.364*** 0.366*** 0.362*** 0.360*** 0.358*** 29 month Partner drunk up till twice a week 2.144*** 2.141*** 2.147*** 2.147*** 2.144*** 2.139*** 2.137*** Partner drunk every day 8.305*** 8.305*** 8.302*** 8.305*** 8.298*** 8.300*** 8.298*** EIGE’s GEI -2.105* Gender Gap Measure -1.498 Overall Population Attainment -1.634*** Measure Women’s abs. achievement -1.430** Men’s abs. achievement -2.291*** Women’s abs. ach. work and -2.161** money domain Men’s abs. ach. work and -1.904** money domain _cons 3.474*** 3.273*** 3.267*** 3.012*** 3.761*** 3.770*** 3.620*** sd country level 0.35 0.38 0.32 0.34 0.31 0.34 0.34 sd individual level 4.13 4.13 4.13 4.13 4.13 4.13 4.13 icc 0.007*** 0.008*** 0.006*** 0.007*** 0.006*** 0.007*** 0.007*** BIC 154832 154836 154827 154830 154826 154830 154830 N 27195 27195 27195 27195 27195 27195 27195 Notes: 1 * p<0.05, ** p<0.01, *** p<0.001 2 Bold variables are country-level variables and their coefficients Source: Authors’ calculations based on the FRA’s survey on Violence Against Women Survey dataset, 2012 and other sources. 30