Violence against women and girls

Summary

This medium investigation seeks to better understand the most promising interventions to reduce levels of violence against women and girls (VAWG). VAWG—mostly acting through intimate partner violence—is common, affecting around 1 in 3 women worldwide, and is associated with a range of negative impacts. These impacts are significant, yet difficult to measure; they span multiple dimensions (wellbeing, physical health, economic) and generations (through spillover impacts to children in the household).

The first aim of this report is to quantify the impacts of VAWG, using existing evidence for VAWG’s impacts. My estimates suggest that most of the burden of VAWG acts through its impacts upon mental health and wellbeing, but there are also substantial impacts to physical and economic health. I also find evidence for impacts to children who witness violence (e.g. who witness intimate partner violence in the household); witnessing violence is stressful, and can negatively impact children’s development. The second aim of this report is to examine whether there are ‘high leverage’ options to reduce VAWG, for example to shift patterns of VAWG in a long term manner. To analyze this, I use four different information sources: theoretical models of VAWG, cross-country comparisons, experimental studies of different interventions, and case studies of VAWG in LMICs. The potential interventions I examine include; (1) increasing rates of land ownership among women; (2) increasing women’s education; (3) increasing women’s income; (4) improving divorce laws such that it is easier to get divorced; (5) reducing alcohol consumption, and; (6) interventions that seek to shift harmful social norms.

A key concern for philanthropists interested in reducing VAWG is that (unlike many well-known and effective interventions such as vitamin A supplementation and bednets) the success of different approaches to VAWG will be highly variable across contexts, in ways that can be difficult to predict. Philanthropists should minimize this risk by strategically attending to known interaction effects among different interventions and risk factors, and by working with groups that have a strong understanding of the local sociocultural context. I include discounts for ‘risk of unintended deleterious effect’ within my estimates of the different impact of potential interventions, but some interventions that I studied are likely to have variable impacts across specific programs (such as mass media).

Following a review of existing evidence, I highlight three specific intervention types which appear especially promising. First are mass media approaches; using edutainment to tackle harmful social norms. Social norms that promote VAWG (e.g. that violence against women is socially acceptable, or that men should have authority over women) are strongly predictive of VAWG rates at a cross-national level. I estimate the impact of a charity using this approach could be around 13X GD. Second is advocacy for changes in land ownership laws—to improve women’s access to land (i.e. through advocacy for marital property rights for women, and to change discriminatory laws where land is preferentially inherited by male heirs), which are associated with lessening women’s ability to leave abusive relationships and consequent bargaining power. I estimate this approach at around 7X GD, when including only the impact upon reducing IPV (in reality, I think this intervention has additional benefits). Third are community activist social empowerment programs; using programs that engage the local community to tackle harmful social norms, and also highlighted in previous work. I estimate that these programs are around 1.8X GD, with considerably less uncertainty than edutainment or land ownership reform. Taken together, I think that there are likely to be opportunities to give cost-effectively to reduce VAWG.

What is VAWG?

VAWG is a broad umbrella term, referring to acts of physical or sexual violence against women; examples include intimate partner violence (IPV), female genital mutilation, honour killings, rape as an instrument of war, sexual trafficking, and the selective abortion of female foetuses. While all these acts of violence are unacceptable, this report focuses upon intimate partner violence—primarily because IPV appears to be especially common, relative to other forms of VAWG (see Fig 1). Note that all forms of VAWG are somewhat interlinked, in the sense that they are thought to share contributing factors (such as the presence of power differences between men and women in society.)

Fig 1: proportional Venn diagram of experiences of violence among 24,000+ women in 15 global sites, taken from Heise (2011).

How do researchers identify and define IPV? Most work in LMIC uses surveys to identify the presence of sexual or physical violence (there is less work looking at emotional abuse in LMIC). For example, Sardinha et al. (2022) found that around ~13% of women globally have experienced physical or sexual IPV within the previous year. The definition of physical IPV was operationalised as ‘acts that can physically hurt the victim, including, but not limited to: being slapped or having something thrown at you that could hurt you; being pushed or shoved; being hit with a fist or something else that could hurt; being kicked, dragged, or beaten up; being choked or burnt on purpose; or being threatened with or actually having a gun, knife, or other weapon used on you; or a combination of these acts’. Sexual IPV was defined as ‘being physically forced to have sexual intercourse when you do not want to; having sexual intercourse out of fear for what your partner might do or through coercion; or being forced to do something sexual that you consider humiliating or degrading; or a combination of these acts’. These definitions were typical of the studies that I came across.

Both men and women can be victims of IPV, but the majority of victims are women; in the UK ~77% of domestic violence victims are women, and ~92% of police-recorded rapes. There is less data available from LMIC, but work by USAID in Ghana and Uganda suggests that (while IPV is common among both men and women) married women were ~2-2.5X as likely as men to have experienced spousal violence, women were more likely to experience all forms of IPV (physical, sexual and emotional) and the violence against women was more severe in terms of likelihood of resulting in injury. Based on the evidence reviewed below (from theoretical models, cross-national correlates, experimental evidence and country specific case-studies) I think it is very likely that power imbalances between men and women play a causative role in the gender imbalance of IPV victimization. This report therefore primarily focuses on IPV against women, under the assumption that the burden of IPV is significantly larger among women than men. Nonetheless, it is important to recognise the burden to male victims. While I have not focused upon this, it is plausible that some of the interventions highlighted within this report may also reduce IPV aimed at men (e.g. interventions that highlight the unacceptability of violence).

Why is VAWG important?

VAWG is widespread; roughly 27% of women who have ever been in a relationship have experienced either physical and/or sexual violence from their intimate partner. The direct impact on women affected is substantial, with the Global Burden of Disease (GBD) identifying intimate partner violence as the 19th leading cause of disease burden worldwide— it accounts for 8.5 million Disability-Adjusted Life Years (DALYs) and approximately 68,500 deaths annually. While the GBD mainly focuses on physical health impacts, VAWG also inflicts mental health and economic impacts. In particular, experiencing VAWG is associated with increased risk of mental health problems such as depression, PTSD and anxiety. VAWG is also associated with lower work wages, through missed days of work. Existing estimates from the Copenhagen Consensus Centre suggest that intimate partner violence costs $66.7 billion per year globally, or 1.4% of world GDP.

Some of these impacts may spillover to other members of the household—such as children who witness domestic violence. In particular, many researchers have argued that witnessing domestic violence is an independent risk factor for various mental health disorders and stress-related chronic illnesses, impaired cognition, and for experiencing relationship violence in adulthood. Most of these impacts are thought to occur through increasing chronic stress levels in early development, and there is strong evidence that early life stress affects brain structure and function. My estimates (below) indicate that these spillover effects are potentially substantial, and of a similar magnitude to the direct impact to the victim (when accrued across all children; in regions in which we might target, women frequently have multiple children).

Quantifying the negative impacts of VAWG, even very roughly, is helpful for philanthropists seeking to work out where the marginal dollar can have the greatest philanthropic impact. While all acts of IPV are moral transgressions that violate the victim’s human rights, there is huge variability in women’s experience of IPV—physical impacts from IPV can range from minor injuries to death— and I note that my estimates seek to capture the average impact. In general, it was very difficult to calculate these estimates due the breadth of impacts that VAWG can have, and the lack of data on VAWG (see more details in the appendix.) Consequently, I have uncertainty in these estimates.

I estimate the impacts of experiencing a year of VAWG in terms of DALYs, WELLBYs and income doublings. I use DALYs to estimate health-based impacts of IPV; one DALY is equivalent to the loss of one year of full health. I use WELLBYs to estimate the wellbeing impacts of IPV; the loss of one WELLBY is equivalent to a 1-point decrease on a 0-10 life satisfaction scale for one year. Finally, I use income doublings to estimate the economic impacts of IPV; we use income doublings to account for our intuition that a given amount of money will matter more to someone on a lower income.

My estimates suggested that being in a relationship that involves IPV for one year causes a total mental health impact of 0.8 WELLBYs; 0.5 WELLBYs to the direct victim during the relationship, and 0.3 WELLBYs to children who witness IPV. I estimated a total physical health impact of 0.1 DALYs to the mother, and an economic impact of 0.02 income doublings. To put this all into perspective, a year’s worth of depression is associated with a per person cost of ~1.3 WELLBYs, implying that the direct experience of being in a relationship that involves violence (my 0.5 WELLBY estimate) is a little less than half the magnitude of depression. The 0.1 DALY burden acts primarily through the physical impact of violence; this is slightly lower than the DALY cost of moderate neck pain (0.114 DALYs), but higher than that of mild diarrhoeal diseases (0.074 DALYs). 0.017 income doublings is equivalent to missing 0.85% of one’s income (£340 of a £40,000 income; we use income doublings to account for our intuition that a given amount of money will matter more to someone on a lower income). These impacts are shown in the figure below, split by direct impacts to the victim versus spillover effects to child witnesses of domestic violence.

Fig 1: Schematic of the impacts I estimated within my BOTEC. Percentages indicate the amount of that impact that comes through that specific factor (so on the far right, 100% of the economic impact to children is coming through the effect of chronic stress upon IQ. This economic burden is 25% of the total impact upon child witnesses, with is 31% of the total VAWG impact. )

What should a philanthropist do to reduce levels of VAWG?

This is a medium level investigation, and as such I still have considerable uncertainty as to the best use of philanthropic money to reduce levels of VAWG: it is possible I have ruled out interventions too quickly, or that there are additional interventions that I have missed out. However, I think that there are three interventions which appear especially promising- I also highlight proposed next steps.

Mass media

The potential for mass media stems from the number of people that an intervention can reach, and the potential to change underlying social norms (which appear causally related to IPV). There is some (small) chance that a mass media program could have longlasting and spillover effects.

The downside is that this is also difficult to empirically test; there is some experimental support for this idea but the existing evidence base is small. I am also skeptical of how informative an RCT of a particular media campaign will be, in assessing the impact of another campaign- I think external generalisability of mass media RCTs is probably fairly low.

A focus for future work should be to understand which mass media programs are likely to be most effective. This will entail understanding the features that are important for success, and the context of particular areas that programs work within (for example, whether the program is tackling the correct social norms in a persuasive manner- the quality of the campaign will be hugely important). Existing NGOs in this area include Population Media Center, Population Foundation of India, Soul City Institute, PCI Media Impact, Uzikwasa and BBC Media Action.

Land ownership reform

The potential for land ownership reform stems from the number of people who would be impacted by reform, which is large, and the potential for longlasting effects (e.g. the potential to change a law that otherwise might not have changed, or not for a long time).

One downside is that this is difficult to empirically test; the existing evidence base does suggest that VAWG will decrease when women own more land, but this evidence base is fairly small. A target for future research will be to better understand the tractability of this intervention, and (if this still looks promising) to ascertain which areas should be focused upon, and how to avert any risk of negative impacts. Importantly, it might be the case that policy changes are unlikely to be actually acted upon in certain areas; for these reasons, I use a low probability of success in my BOTEC (yet it still comes out as fairly cost-effective).

An additional focus for future work should be to understand the additionality of funding land ownership reform; I am uncertain of RFF for NGOs focused upon this. I think it would also be beneficial to spend some time tracking down NGOs in this area, including local NGOs who may be less google-able. An existing NGO that I am aware of is Landesa.

Community Activist Social Empowerment

The potential from CASE programs stems from their efforts to shift social norms in a locally-driven manner, and the fact that these programs have a strong evidence base behind them. It is possible that while CASE programs work, mass media programs (which ultimately function in a similar way, by trying to shift social norms, but are lighter touch) might not.

The downside is that it seems unlikely that they are extremely cost effective (>10X GD) due to their resource intensive nature; they can reach a far smaller number of people per donation amount, and may be difficult to scale.

A focus for future work should be to understand the RFF of CASE programs. It is also possible that I have been too skeptical in assuming that these programs are not >10X GD, for example if there is evidence of spillover or longlasting effects—but my understanding is that there is no evidence for this at present.

Existing NGOs include Raising Voices, and CEDOVIP.

Conclusions

IPV is common, and has severe impacts (for example, with regards to wellbeing, economic status, and women’s empowerment more broadly). These impacts are diffuse and difficult to measure, spanning different areas of womens’ lives, are likely to impact children in the household, and may have effects on economic growth.

To the best of my current understanding, risk-neutral philanthropists should favour programs such as large scale mass-media approaches that aim to shift harmful norms around VAWG, and advocacy approaches that aim to improve women’s land ownership. These are high-leverage options, which are backed by empirical work. Future information that could shift my view around this are if there is less room for funding in these areas than I anticipate, if advocacy over land rights is far less tractable than I am assuming, or if methodologically stronger work than currently available suggests that the relationship between land ownership and IPV is not causal. I think there is a lot to be gained from studying these two approaches in more detail, to work out which organizations within these spaces are likely to be most impactful—this will entail significant knowledge of the local cultural context, perhaps especially for mass media since it focuses on social norms. Finally, community Activist Social Empowerment programs are also likely to be somewhat cost effective (but below the 10X GD mark) and are less risky.