Measuring violence, mainstreaming gender: does adding harm make a difference?

465 Journal of Gender-Based Violence • vol 2 • no 3 • 465–79 • © Centre for Gender and Violence Research 2018 University of Bristol 2018 • Print ISSN 2398-6808 • Online ISSN 2398-6816 https://doi.org/10.1332/239868018X15366982109807 This article is distributed under the terms of the Creative Commons AttributionNonCommercial 4.0 license (http://creativecommons.org/licenses/by-nc/4.0/) which permits adaptation, alteration, reproduction and distribution for non-commercial use, without further permission provided the original work is attributed. The derivative works do not need to be licensed on the same terms.


Introduction
How we define and measure violence has implications for understandings of violence as well as for policies to stop violence. In several publications, Sylvia Walby, Jude Towers and colleagues (Walby and Towers, 2017;Walby et al, 2014Walby et al, , 2016Walby et al, , 2017 have criticised current statistical practices and survey methodologies for underestimating the gendered patterns of violence, as well as for a lack of consistency over time and across studies and countries. In recent publications they criticise the design of surveys on intimate partner violence (IPV), which rely on the conflict tactics scale (CTS) (or modified versions of it), an American survey instrument originally developed by Gelles and Straus in 1979, which counts acts alone. Indeed, studies using this scale have been criticised for failing to contextualise acts defined as violence (Dobash et al, 1992). According to Walby and Towers (2017) and Walby et al (2017), however, the main problem with CTS and its modified versions is that: 'Actions alone are not sufficient to define either violence or crime' (Walby et al, 2017, 33), as gender mediates the consequences of acts, and the same act may have different consequences if carried out by a man towards a woman and vice versa. These possible gender differences are not captured in studies that assume that the same act always represents the same level of violence.
Based on their critiques of current methods of measurement, Walby and Towers propose a set of criteria for mainstreaming gender in surveys on IPV, including several modifications to the survey methodology such as new operationalisations of 'violence' and 'gender' to 'assess the distribution of violence including its multiple gender dimensions ' (2017, 12).
As part of harmonising population surveys of violence with other statistics as well as across countries, Walby et al (2017,36) suggest anchoring surveys on violence in international and national law, and bringing the definition of violence in surveys in accordance with the definition of violence in law, arguing that 'actions (and intentions) and harm (and nonconsent) are together necessary for the definition of violence for use in the measurement framework'.
This article takes as its starting point Walby and colleagues' proposals for a new operationalisation of violence and gender in survey studies. Based on the criteria Walby and Towers suggest, it analyses the effectiveness of the most recent Norwegian national survey on violence and rape among the adult population (Thoresen and Hjemdal, 2014) in revealing gendered patterns of exposure to violence, and the ways in which a partial adoption of Walby and Towers' operationalisation of violence affects the gendered distribution of violence in this study. This investigation represents a suitable case for testing the possible benefits of adopting their suggested framework and definition of violence, as on the one hand the study belongs to the kind of studies that Walby et al criticise, given that it measures violence as acts alone, but on the other hand it fulfils to some extent the proposed criteria for mainstreaming gender in surveys. Furthermore, it also contains data on harm for acts defined as severe violence and sexual violence. This will enable a re-analysis of the data, in order to test whether the application of Walby and Towers' definition affects the gender distribution of exposure to severe physical violence and sexual violence. The aim of this article is to contribute to the ongoing debates on how to improve survey methodologies to obtain more accurate knowledge of the gendered distribution of exposure to violence, and to facilitate greater comparability with other studies over time and in different countries.
The article will first briefly present Walby and Towers' proposal, and then assess how effectively the Norwegian survey performs according to their proposed criteria for mainstreaming gender and defining violence. The authors will then present the gender distribution of violence, with specific attention to intimate partner violence based on the findings from the survey, using the original definition of violence as acts alone. The article subsequently presents the results of a re-analysis of the survey data applying Walby and Towers' proposal, adding harm to the definition of violence, and discusses the implications of adding measures of physical harm and fear of being severely injured or killed.

Definition of violence
Walby and Towers suggest a new and narrower definition of violence that includes actions and harm (and intentions and non-consent).

Mainstreaming gender
In order to to mainstream gender into surveys and to more effectively capture possible gendered patterns of violence, Walby and Towers (2017, 13) suggest that: At least four gender dimensions should be included, so that the 'gendersaturated' nature of violence can be addressed. First is the sex of the victim; second is the sex of the perpetrator; third is the relationship between perpetrator and victim (intimate partner or other family member; acquaintance; or stranger); and fourth is whether there is a sexual aspect. A further, possible, fifth dimension is whether there is an identifiable gender motivation.

The Norwegian survey
The Norwegian survey on violence and sexual abuse was conducted with a representative sample of the Norwegian population in 2013 ( Thoresen and Hjemdal, 2014). The survey used selected questions from the Conflicts Tactics Scale (Straus et al, 1996), the National Women's Study (Kilpatrick et al, 1992) and the National Violence Against Women Survey (Tjaden and Thoennes, 2000) (see Thoresen and Hjemdal, 2014). It covered a range of experiences of physical violence over the life course, from childhood to the present, as well as questions of sexual abuse, coercive control, stalking and measures of mental health and feelings of shame and guilt.
A representative sample of 40,000 men and women between the ages of 18 and 75 years was drawn from the Norwegian population register and invited to take part in the study. The study was conducted as a computer assisted telephone interview (CATI) study and the interviewers were able to reach a total of 9,647 people on telephone. Of these, 5,120 people declined to answer the questions and 2,435 women and 2,092 men were interviewed, which gives a response rate of 46.7 per cent of those answering the telephone (the total population in Norway was approximately five million in 2013). Prevalence data were weighted for gender, age and area of residence. The weights were constructed as inverse probability weights for the sample of responders based on population figures from Statistics, Norway. Only minor differences were found between weighted and unweighted prevalences. Gender differences in violence exposure were tested with chi square statistics. All analyses were conducted in SPSS for Windows version 25.
We analyse below how well the Norwegian survey performs according to the quality criteria proposed by Walby and Towers (2017): see Table 1. The Norwegian survey shares many characteristics with the European survey conducted by the European Union Agency for Fundamental Rights in 2014 (FRA), which also appeared in Walby and Towers' article as an example of the genre violence against women studies in contemporary European studies, and we have retained the FRA study as well as those from the Office for National Statistics (ONS) and Crime Survey for England and Wales (CSEW), in our adapted table for comparison. In line with Walby and Towers' distinction between published and potential data from the ONS and CSEW, we also distinguish between the published data from the Norwegian study and the data that can potentially be obtained from it, and in the analysis to follow we draw on the published data from the survey as well as on a re-analysis of the data.

Anchor of study
The Norwegian survey is anchored in the World Health Organization's (WHO) definition of violence (Thoresen and Hjemdal, 2014). This has no reference to legal frameworks, and the definition of violence is not rooted in the law, with the exception of sexual violence, the definition of which coincides with legal definitions. Its anchoring in a health frame rather than a legal frame is also reflected in the emphasis in the questionnaire on mental health.
The anchoring of the study can be extended to additionally include national and international law, albeit not in full. The Norwegian Penal Code's distinction between inflicting bodily harm and assault is not based on wether the violence results in physical harm or not, but rather whether the injury represents a direct and intended consequence of the act or not. Thus, a respondent may have suffered physical damage as a consequence of the violence, but it was nevertheless an 'assault' (injurious), rather than 'inflicting bodily harm' as the harm was not intended; for example, a blow causing the victim to step backward, stumble and hit his or her head.

Definition of violence
Violence is defined in the survey as acts alone, distinguishing between severe and less severe acts. Less severe violence was measured by four items based on the Conflicts Tactics Scale (Straus et al, 1996): hit you with an open hand; pulled you by the hair; scratched you; pinched you hard. Severe violence was measured by six items based on the National Survey of Adolescents (Kilpatrick et al, 2003): hit you with the fist or a hard object; kicked you; choked you; beat you up; threatened you with a weapon; assaulted you physically in other ways. The study includes separate questions on rape and sexual violence. Rape is measured by four items based on the National Women's Study (Kilpatrick et al, 1992): forced you to have intercourse by using physical force or threatening to hurt you or someone close to you; forced you to have oral sex by using physical force or threatening to hurt you or someone close to you; forced you to have anal sex by using physical force or threatening to hurt you or someone close to you; put fingers or objects into your vagina or anus by using physical force or threatening to hurt you or someone close to you. Sexual abuse was measured by the following items based on the National Survey of Adolescents (Kilpatrick et al, 2003): unwanted sexual contact when you were too intoxicated to consent or could not stop it from happening; touched your genitals or made you touch their genitals by using force or threatening to hurt you; pressured to perform sexual acts (this case implies penetration of your vagina/oral sex or anal sex); other forms of sexual offences or abuse (Thoresen and Hjemdal, 2014). Both acts and harm are measured, but questions of acts and harm are asked separately, and it is not possible for all questions to link particular acts and harm to particular perpetrators if more than one perpetrator exists. In addition to the question about physical harm the respondents are also asked about fear of being severly inured or killed by the violence or rape. Follow-up questions on harm and fear are only asked in relation to severe physical violence and forced rape. Due to an oversight during the construction of the questionnaire, no question regarding intoxicated rape for adults over the age of 18 was asked. In addition to questions of injury pertaining to severe violence and sexual violence, questions regarding psychological reactions (anxiety, depression and post-traumatic stress) and feelings of guilt and shame are asked in a separate section.
Allthough the published data rely on the definition of violence as acts alone, the definition may be modified to include harm in line with Walby and Towers' proposal.

Gender
The survey applies both to men and women, and identifies the gender of the victim, the gender of the perpetrator and the relation between victim and perpetrator. In addition, the study contains questions on rape and sexual abuse. The study subsequently comprises the main gender dimensions suggested by Walby and Towers.

Number of events/repetition
The number of events is grouped, with six or more incidents as the highest possible number for each type of physical violent act, and 11 or more incidents as the highest possible number of rapes and other forms of sexual abuse. Like the FRA study, the Norwegian study therefore does not count all incidents and so only partially satisfies Walby et al's quality criteria on this point.

Non-consent/intention/gender motivation
Non-consent is included for sexual violence only. Intention is excluded. Gender motivation is also excluded. In Norway, the ethical research guidelines substantially limits the possibilities of asking questions about third persons and it is thus not possible to ask victims about the perpetrator's motivation or intent. Gender motivation as suggested by Walby and Towers is not covered, and it is difficult to see how it could be operationalised in a Norwegian context. Summing up, the Norwegian survey performs rather well on the gender dimensions suggested by Walby and Towers, and although it only partially delivers on the other dimensions, it has the potential to be useful for a systematic gender analysis based on the published data. Furthermore, given that the study contains questions for both acts and harm, it is also possible to undertake a re-analysis of the data, using Walby and Towers' proposed definition of violence to include harm. In the following section, we first present an analysis of the gender dimensions of the study based on the published data, before sharing a re-analysis using Walby and Towers' definition (adding harm, as well as fear of being injured or killed) in order to explore its effects on the gender pattern of exposure to violence.

Violence and gender in the Norwegian survey
In the following section we will present the gender distribution of violence, based on data from the survey, using its original operationalisation of violence as acts alone, and predominantly drawing on the published data (Thoresen and Hjemdal, 2014).
Over the life course, 52.9 per cent of women and 58.0 per cent of men reported at least one experience of physical violence and/or sexual abuse, whether as adults or chidren. Exposure to severe physical violence after 18 years stands at 22.5 per cent for women and 44.5 per cent for men, and so in general, men are more exposed to severe physical violence than are women. Among victims of all forms of severe violence over the life course, women feared being seriously injured or killed to a greater extent than men (49.6 per cent and 31.4 per cent respectively), while the share of victims who were physically injured was rather similar (55.7 per cent of women as opposed to 53.1 per cent of men) (Thoresen and Hjemdal, 2014, 77-9).
As regards sexual violence, women are more at risk: over the life course 9.4 per cent of women and 1.1 per cent of men had been raped, and lifetime exposure to all sexual abuse stands at 33.6 per cent for women as opposed to 11.3 per cent for men. Among victims of rape over the life course, women were much more likely to report being physically injured (28.8 per cent as opposed to 4.3 per cent of men), and to a greater extent feared being killed or injured (44.4 per cent of female rape victims compared with 37.5 per cent of male victims) ( Thoresen and Hjemdal, 2014, 85).
For all types of violence aside from partner violence, the majority of perpetrators, whether against men or women, were men. Men were the perpetrators in almost all cases of rape against women (97.8 per cent of 229 cases), and in 15 of the 24 rapes against men. In six cases a woman was the perpetrator, and in three cases of rape against men, the perpetrators included a woman and a man (Thoresen and Hjemdal, 2014, 89).

Relation to perperator
For women and men who are exposed to violence, their relationship to the perpetrator differs. Congruent with previous studies, the most common perpetrator for women exposed to severe violence was a partner or previous partner (40.9 per cent, in comparison to the 4.3 per cent of male victims who faced severe violence from a partner). In contrast, men were far more likely to experience violence from a stranger: for 72.1 per cent of men exposed to severe violence the perpetrator was a stranger, as compared to 26.1 per cent for women (Thoresen and Hjemdal, 2014, 77).
In the following section we look more closely at partner violence, and, based on the findings of the survey we show how partner violence can be presented as both gender symmetrical and as gendered.

Claim I: IPV is gender symmetrical
If we consider all kinds of physical partner violence (severe and less severe physical acts, excluding sexual violence), 17.1 per cent of women and 17.2 per cent of men have encountered some form of physical violence from a partner at least once during the life course. For less severe violence from a partner, gender symmetry can also be identified: slightly more men (16.3 per cent), than women (14.4 per cent), have had experiences of less severe violence from a partner (Thoresen and Hjemdal, 2014, 79): see Table 2.

Claim II: IPV is gendered
If we consider severe partner violence, however, the picture is gendered, as summarised in Table 3: 9.2 per cent of the women and 1.9 per cent of the men reported severe physical violence from a partner during the life course. Women were also exposed to a higher number of severe incidents from a partner than were men. Due to the design of the questionnaire, it is not possible to distinguish repeated severe violence from a partner, from repeated severe violence from other perpetrators (Thoresen and Hjemdal, 2014, 121).
If we consider sexual violence from a partner, the picture is even more gendered: 3.8 per cent of women and 0.1 per cent of men had been raped by a partner. In addition, 5.5 per cent of women and 0.5 per cent of men had been exposed to other forms of sexual abuse from a partner (Thoresen and Hjemdal, 2014, 81).
If we look at polyvictimisation, more women are exposed to multiple forms of severe violence and sexual abuse: 7 per cent of women and 0.8 per cent of men experienced at least two of the following categories of severe violence: severe physical violence where the victim feared being injured and killed; severe partner violence; rape (Thoresen and Hjemdal, 2014, 89). Women experienced a higher number of events: 40 per cent of the women who had experienced less severe violence had experienced more than five events, compared to 28 per cent of the men (Thoresen and Hjemdal, 2014, 80). However, this does not offer a full picture of multivictimisation and how it might be gendered, as we do not have access to the actual number of incidents, owing to the scale used in the survey in which 'more than five times' is the highest number of incidents possible to report for each of the four specified kind of acts.
In conclusion, the Norwegian survey in its original form, defining violence as acts alone reveals a picture of violence as highly gendered: women and men are exposed to different forms of violence and from different perpetrators, and women are much more exposed to severe partner violence, to a greater amount of violence, and to multiple forms of severe violence. Furthermore, women are more exposed to all forms of sexual violence, and sexual violence from a partner is almost exclusively experienced by women. The report from the survey also concluded that violence is a gendered phenomenon (Thoresen and Hjemdal, 2014, 129).
Intimate partner violence can, however, also be presented as gender symmetrical, albeit only if using a broad definition of violence as acts alone, including less severe acts and excluding sexual violence, polyvictimisation and frequency. Table 4 presents an overview of the two different approaches. Although it is possible to present the data in such a way, doing so conceals the gendered pattern of exposure, and the survey does not support a general claim of gender symmetry in exposure to violence.
The means and order of presenting data is not insignificant. Indeed, it has important epistemological and political implications: what we claim to know and in what order we present that knowledge matters. Starting with and thus foregrounding the most severe violence shifts the focus towards the violence that has the most serious impact, and also towards the violence that is closer to the criminal definition. Starting with gender symmetry, in contrast, may implicitly convey the impression that this constitutes the main pattern, reducing gender differences to a secondary and less important subset of interpretations. The order and implicit hierachisation of knowledge may again have implications for how it is received and understood, and for how numbers are invoked in struggles over recognition, resources and policies. Adding harm: what happens to the gender distribution of violence?
In the following section we present a re-analysis of the survey data, employing Walby and Towers' definition of violence, including harm, and we also incorporate fear of being seriously injured or killed. We define harm as physical injury as a direct consequence of the violence at the time when it happened. We can only include physical injury for severe violence and rape, as no questions regarding physical injuries related to less severe violence and forms of sexual abuse other than rape were asked using the survey. The following analysis is limited to events after the age of 18, as we only have comprehensive data for these cases. First, we look at how the gender distribution changes for the total sample if we add injury and fear. The measure for gender distribution is computed as a simple ratio by dividing the prevalence rate for women with the prevalence rate for men.
As we can see (Table 5), adding injury hardly changes the gender distribution for severe violence after the age of 18. Twice as many men were exposed to severe violence, and adding injury scarcely changes the gender ratio from 0.5 to 0.53. Adding fear of being killed, however changes the gender distribution, illustrating that more women than men exposed to severe violence feared being injured or killed.
As we can see (Table 6), rape is heavily gendered, with more than 20 times more women having been raped after the age of 18. Among the women who had been raped, one third had been injured, while half feared being injured or killed. None of the men were injured, but one third feared being injured or killed. The number of men is so low, however, that these figures should be treated with caution.
We will next look at partner violence. In order to distinguish violence from a partner and violence from other perpetrators, we present data both for those who had been exposed to violence from a partner (and potentially also from others) and those who reported violence or sexual violence from a partner only.  As we can see (Table 7), the gender distribution for severe violence hardly changes when we add injury, but the gender difference increases when we add fear of being injured or killed. For those exposed to severe violence from partner only, injury also makes a difference, changing the gender ratio from 7.0 to 9.3, but this is not significant. Adding fear of being injured or killed, however, changes the gender ratio to 17.9, which is significant (p<0.05), using the bootstrap method owing to low numbers. Fear of being injured or killed seems to be particularly important for those who have been exposed to severe violence only from partners. This may indicate that the gender direction of partner violence is important in terms of the perceived threat it represents, and that men to a lesser extent react with fear to partner violence, compared to women: only a quarter of the exposed men were afraid, whereas almost two thirds (61 per cent) of the exposed women feared being injured or killed.
As the survey also includes questions about controlling behaviour, and we have the possibility to look into whether the violence also includes controlling behaviour, and how this affects the gender distribution. The control items build on the division suggested by Michael Johnson (1995) between common couple violence and patriarchial terrorism (later renamed intimate terrorism). Johnson operationalises intimate terrorism as all physical violence towards a partner if the partner is also subjected to at least three out of seven different forms of controlling behaviours (Johnson and Leone, 2005). As the survey includes only three of Johnson's questions, we have chosen to define the violence as intimate terrorism only if all three forms of controlling behaviour are present. Including this measure does not, however, change the gender distribution of physical parter violence, and women and men who were exposed to phsycal violence were equally exposed to controlling behaviours by a partner.  Finally, we consider rape by a partner (Table 8). As none of the male victims were injured and none feared being injured or killed, we were unable to calculate a gender ratio.
Among the female victims, slightly more than one third were injured, while more than half feared being injured or killed. The gender distribution was almost the same for those who reported rape from a partner and those who reported rape from a partner alone.

Adding measures for mental health
The study also included two instruments measuring mental health: the Hopkins Symptoms Checklist (HSCL) measuring depression and anxiety and the Posttraumatic Stress Disorder Checklist (PCL). Adding the scores for depression and anxiety and for post-traumatic stress reveals that all violence has an effect on mental health. Adding injury in itself does not significantly change the scores on mental health for men, and only slightly for women. Fear of being injured or killed, in contrast, has a significant impact on mental health, both for women and men. Keeping in mind that an intimate relation to the perpetrator increases the level of mental health consequences, and that for women an intimate partner is the largest group of perpetrators, and further, that a larger share of women expressed fear of being injured or killed, the number of women affected is thus higher. The men who expressed fear in relation to severe violence, however, had a similar increase in mental health scores as the women.

Discussion
Adding harm in line with Walby and Towers' proposal did not change the gender ratio of exposure to severe physical violence and sexual violence in this study. Rather, acts categorised as severe physical violence and sexual violence appeared to represent an adequate measure of violence, which revealed gendered patterns of exposure. We do not have access to data on harm pertaining to less severe violence, so we cannot generalise to all acts defined as violence in the Norwegian survey. Adding fear of being injured or killed significantly changed the gender ratio, however. Consequently, rather than adding injury, adding fear might be important in a gender-sensitive measure of violence. In their proposed questionnaire, Walby and Towers include an item for threats, but not fear. However, being threatened is not the same as feeling fear, and fear may also occur without threats being explicitly formulated.
In the Norwegian study, adding injury significantly reduces the number of victims of violence, halving it for women and more than halving it for men. A risk exists in adopting a narrower definition of violence as act + harm, as victims of severe violence who may have been exposed to criminal acts and who may suffer serious health impacts would be excluded from the statistics. Adding injury to the definition of sexual violence would also lead to an underestimation of sexual violence, because physical injury is not necessarily linked to fear or psychological reactions.
Defining violence narrowly risks underestimating the number of victims who are seriously affected and who may be at risk of being harmed or killed. Fear of being injured or killed may provide important and life-saving information as to the threat potential of the violence, as victims' perceived risk is increasingly recognised as important in assessing risk of homicide (Sharp-Jeffs and Kelly, 2016) and/or in need of help.
Finally, one important motivation for Walby and Towers' proposal to change the definition of violence to include injury was to harmonise the definition of violence and sexual violence in studies of violence with the legal definitions, thus facilitating comparison across different studies as well as across countries. However, this might be difficult, as the legal definitions differ between countries, as well as within Europe.

Conclusion
We fully sympathise with Walby and colleagues' endeavours to improve methods of measuring violence and to mainstream gender in studies of violence as part of ending violence. In this article, we have demonstrated that the Norwegian survey on violence and rape to a certain extent fulfils the criteria for gender mainstreaming launched by Walby and Towers, as it helps to account for multiple gender dimensions and for gender differences in exposure and effects of violence. The analysis has also shown that certain aspects of the Norwegian survey could be improved, such as counting all events rather than grouping numbers of events, making sure that it is possible to link all acts and repetition to particular perpetrators, including when there are multiple perpetrators, and to add questions of harm related to less severe violence. The analysis also illustrates how gendered patterns may be foregrounded or downplayed in the presentation of the findings.
Based on the analysis and re-analysis of the data in the survey, however, we do not find support for adding injury to the definition of violence as a means of obtaining a more accurate picture of gender patterns of exposure. Indeed, this study revealed that in the case of severe partner violence and sexual violence, adding injury did not change the gender ratio of victimisation. For both men and women, however, adding injury reduced the number of victims of violence, hence using a definition of violence that includes harm reduces the general prevalence. We are concerned that using a narrower definition might lead to an underestimation of severe violence and sexual violence. Rather than injury, we found that fear of being injured or killed made a gender difference.