CGD NOTES

Violence Against Women and Children During COVID-19—One Year On and 100 Papers In: A Fourth Research Round Up

A year after the World Health Organization declared COVID-19 a pandemic, we take stock of an increasingly diverse set of new studies linking violence against women and children (VAW/C) to COVID-19 and associated pandemic response measures. In this fourth round up, we focus exclusively on research in low- and middle-income countries (LICs and MICs) published since December 2020 to highlight dynamics in settings that previously had fewer studies. As in previous round ups (see the first, second and third covering a total of 74 studies), we only include studies that have sufficient information on indicator definition and analysis methods (though we maintain a full set of studies including more policy-oriented briefs and reports—in our evidence tracker). In total, we summarize 26 new studies from LICs and MICs, with the majority focused on identifying trends (15 studies), while others present analysis of risk factors or dynamics (an additional ten studies), and one represents an impact analysis of prevention programming.

What have we learned? Key takeaways for this roundup

  • More papers point to increases in violence. Of the 15 studies presenting findings on trends in VAW/C from pre-pandemic through various stages of the pandemic, 12 of these papers (or 80%) find exclusive evidence of increased violence—a higher proportion than previous round ups. Of the remaining studies in this round up, one paper finds mixed results, and two find no significant change in violence.

  • Lost income and employment, among other factors, increase the risk of violence occurring. Of the additional ten studies assessing risk factors, findings indicate that being married, unemployed (for either the respondent or spouse), reporting lost household income, food insecurity, and spousal substance abuse tendencies increase risk of VAW/C. Protective factors identified include higher education level for either the wife or husband and women’s employment.

  • As an ongoing issue, too few papers evaluate ‘what works’ to reduce the risk of violence or support survivors. Only one study in this round up provided evidence on intervention effects during COVID-19: a study from Bolivia showed a youth empowerment program decreased multiple types of violence experienced by adolescent girls. This study adds to only a handful of global studies able to inform VAW/C prevention policy and programming. More evidence is critically needed.

  • Studies reflect findings across diverse settings and populations. In this round up, nine studies were from Asia (five from South Asia), eight from the Middle East, six from sub-Saharan Africa, one from Latin America, and two were cross-regional studies. Two studies focused on violence against healthcare workers and two studies focused on pregnant women specifically, all documenting high rates of violence during the pandemic.

An overview of new studies: Research questions, geographies, and methods

Similar to previous round ups, studies continue to focus on the research question of whether or not VAW/C is increasing or decreasing during COVID-19. Of the 15 studies included here, 12 find evidence of increased violence. This represents a higher proportion of studies supporting exclusive increases in violence than in previous round ups (80%, compared to 25% in the first round up, 56% in the second, and 50% in the third round up). Some of these increases are large, for example, studies in Jordan (showing a quadrupling of self-reported violence against women) and China (showing a 278% increase in police calls for domestic violence). Of the remaining three studies, one finds mixed results (in India), and two find no significant change in violence (Kenya and South Africa).

One factor that could explain this difference is the type of data being used. Studies in the three previous round ups mostly relied on administrative data such as calls to emergency services or clinical data from hospital admissions. This focus on ‘reported’ data was likely partially due to data availability in the early stages of the pandemic, concentrated in high income countries, particularly the United States and across Europe.

In this round up, more studies use survey data, collected either online or via mobile phone (11 of the 15 included papers). However, most of these studies rely on small samples that cannot be generalized to larger populations, and many collect data at a single point in time asking respondents to recall experiences from before the pandemic (or use innovative indirect measures, including vignettes , as done in Indonesia). Survey data may capture a greater range of experiences, not only those which are most severe or where help-seeking occurs. While administrative data can be collected and reported on very quickly, many of the studies in this round up use survey data to provide a more comprehensive picture of violence occurring, extending beyond incidents captured in police reports or hotline calls.

Similar to our last round up, several efforts use ‘big data’—including cross country analysis of internet search data, or social media posts. For example, an analysis of 11 countries using Google search data suggests domestic violence increased more in places with stricter lockdowns, as measured by Google mobility data. In addition, an 8-country analysis of social media data in Asia unpacks dynamics in the public sphere, and shows online help-seeking increased (from 10 to 70%) across all but two countries.

Group A. Papers that measure impacts of COVID-19 or associated response measures on VAW/C

Authors Location Data Methods Indicator(s) Finding
1. Abuhammad 2020 Jordan Survey data (online) Descriptive statistics; multivariate regression Violence against women Increase
2. Aolymat 2021 Jordan Survey data (online) Self-reported trends (descriptive) Domestic violence Increase
3. Berniell & Facchini 2020 Argentina; Brazil; Chile; Colombia; France; Germany; Italy; Mexico; Spain United Kingdom; United States Google search data; google mobility data Event study analysis; difference-in-difference Domestic violence Increase
4. Dai et al. 2021 Hubei, China Police service calls Descriptive statistics; local regression; one-way ANOVA; ARIMA models Domestic violence Increase
5. Fabbri et al. 2020 Nigeria; Mongolia; Suriname Survey data (face-to-face) Multivariate regression Violent discipline Increase
6. Fereidooni et al. 2021 Isfahan, Iran Survey data (face-to-face pre-pandemic and phone during pandemic) Comparison of means during COVID-19 to prepandemic (January and February 2020) levels; multivariate regression Intimate partner violence (IPV) Increase
7. Guglielmi et al. 2020 Bangladesh Survey data (phone); qualitative interviews (telephone) Self-reported trends (descriptive); thematic analysis Gender-based violence; police violence Increase
8. Halim et al. 2020 Indonesia Survey data (phone) Self-report trends (descriptive); machine learning IPV, violence against children, harassment (all reported “in the community”) Increase
9. Mahmood et al. 2021 Kurdistan region, Iraq Survey data (online) Self-reported trends (descriptive) Spousal violence Increase
10. Pattojoshi et al. 2020 India Survey data (online) Self-reported trends (descriptive) Spousal violence (physical, sexual, verbal, emotional) Increase
11. Pinchoff et al. 2021 Nairobi, Kenya Survey data (phone) Descriptive statistics; multivariate regression Household violence; violence outside the home Increase
12. UNFPA et al. 2021 Bangladesh, India, Indonesia, Malaysia, Nepal, Thailand, Philippines, Singapore Social media data; Internet search data Social media discourse analysis; temporal analysis Violence against women Increase
13. Sharma & Khokhar 2021 India Survey data (online) Self-reported trends (descriptive) Domestic violence Mixed
14. Egger et al. 2021 Kenya Survey data (phone) Comparison of means during COVID-19 to prepandemic (March 2020) levels IPV; violence against children No change
15. Venter et al. 2020 Johannesburg, South Africa Hospital clinical assessments Annual comparison of cases Interpersonal violence; trauma cases No change

Table notes: ANCOVA = Analysis of covariance; ARIMA = Autoregressive integrated moving average; IPV = intimate partner violence.

More details on papers that measure impacts of COVID-19 or associated measures on VAW/C

Abuhammad 2020.

Using an online survey of 687 women, Abuhammad shows that the proportion experiencing violence quadrupled during the pandemic (from 10% to 40%), with parents (both mothers and fathers) being the most often reported perpetrators. Less than half of the women who had experienced violence reported it to the police, and perpetrators were arrested in only 3.5% of cases. The most significant predictors of experiencing violence during this period were a woman’s marital status and unemployment, highlighting the link between economic insecurity and violence [Abu-Hammad 2020; International Journal of Clinical Practice.

Aolymat 2021

In a study of the impact of COVID-19 on domestic violence and reproductive health in Jordan, Aolymat finds that 20.5% of 200 women surveyed reported experiencing increased domestic abuse during the pandemic. The author hypothesizes that increased abuse is due to increased time spent at home with partners, reduced income, and reduced access to healthcare services as a result of the pandemic [Aolymat 2021; American Journal of Tropical Medicine and Hygiene].

Burniell and Facchini 2020

Using Google search data on domestic violence-related terms as a correlate for incidence of violence, Berniell and Facchini estimate that incidence of violence peaks at seven weeks into lockdown measures and remains statistically significant until ten weeks into lockdown. Relative to the week prior to lockdown, domestic violence searches increased by 31%. Though an increase was observed in every country in the sample, the effect in Latin American countries was only half that in high-income countries, a difference that the authors attribute to differential adherence to stay at home measures [Burniell and Facchini 2020; CEDLAS Working Paper].

Dai et al. 2021

Using police service call data from the Hubei province of China, the site of one of the strictest lockdowns of the pandemic, Dai, Han, and Xia find that although overall calls to police decreased during the lockdown period, the average number of calls related to domestic violence nearly quadrupled during the lockdown period (an increase of 278% in adjusted models). While most other types of police calls returned to normal levels shortly after the lockdown ended, domestic violence calls remained elevated and took longer to return to pre-lockdown levels [Dai et al. 2021; Policing].

Fabbri et al. 2020

Fabbri et al. model the effects of lockdowns on violent discipline of children in three countries (Nigeria, Mongolia and Suriname) using data from (pre-pandemic) Multiple Indicator Cluster Surveys. They find models predict large increases (35% to 46%) in violent discipline scores under ‘high restriction’ scenarios and smaller increases (4% to 6%) under ‘low restriction’ scenarios [Fabbri et al. 2020; Child Abuse and Neglect].

Fereidooni et al. 2021

Using survey data of 2,116 adult partnered women in Iran, Fereidooni et al. show that prevalence of IPV during COVID-19 rose from pre-pandemic levels (from 54% to 65%) and that over a quarter of women reported first-time incidence of IPV during COVID-19. The authors find that women’s engagement in paid employment decreases the likelihood of exposure to IPV [Fereidooni et al. 2021; SSRN Working Paper].

Guglielmi et al. 2020

In a study of the impact of COVID-19 on Rohingya and Bangladeshi adolescents (1,761 phone surveys and 30 qualitative interviews), Guglielmi et al. find that 8% of adolescents surveyed (boys and girls) reported an increase in gender-based violence during the pandemic, and that about a third of boys and a fifth of girls living in camps reported escalated police and military violence to enforce containment measures. Married girls were twice as likely as unmarried girls to report an increase in gender-based violence in the community [Guglielmi et al. 2020; Journal of Migration and Health].

Halim et al. 2020

Using a phone survey of 866 women in Indonesia and indirect measures (vignettes asking respondents to report on violence “in the community”), Halim et al. show respondents report increases due to COVID-19 of 83% (IPV), 68% (violence against children) and 65% (harassment). Food insecurity and women’s lack of access to jobs are significant correlates of reported increases [Halim et al. 2020; World Bank Technical Brief].

Mahmood et al. 2021

Using online survey data of 346 married women, Mahmood et al. find a significant increase in spousal violence when comparing self-reported instances before and after the lockdown in Iraq. For any violence, the prevalence rate rose from 32% to 39%, and significant increases were also seen for emotional abuse, physical violence, humiliation, intimidation, hitting, hair and arm pulling, and forced sexual intercourse [Mahmood et al. 2021; Journal of Interpersonal Violence].

Pattojoshi et al. 2020

Using an online survey of 560 women, Pattojoshi et al. report a prevalence rate of spousal violence of 18.1%, of which verbal and emotional violence were the most common, followed by physical and sexual violence. About 5% of women reported experiencing violence for the first time since lockdown began, and of those who reported having experienced it before, 78% reported an increase since lockdown. The most commonly perceived reasons for violence were financial constraints, inability to socialize, and sharing responsibilities for childcare [Pattojoshi et al. 2020; Psychiatry and Clinical Neurosciences

Pinchoff et al. 2021

Using data collected from 2,009 households in informal settlements in Nairobi, Kenya, Pinchoff et al. document reported increases in violence against women inside and outside the home (45% and 24%, respectively). Women are 8 percentage points more likely to report increased risk of household violence (as compared to men), particularly in households with higher food insecurity [Pinchoff et al. 2021; BMJ Global Health].

UNFPA et al. 2021

Using social media and internet search data, the authors document the most searched for terms related to GBV across several Asian countries; Singapore, Malaysia, and the Philippines show a marked increase in GBV-related searches between early spring and mid-summer 2020. Online misogyny increased during lockdown in all countries examined, but online support for survivors and services also increased. Help-seeking online increased in most countries by between 10 and 70% [UNFPA et al. 2021; Technical report].

Sharma and Khokhar 2021

In a sample of rural households in Kenya, Eggers et al. find an increase in violence against both women and children during the crisis period (of 4% and 13%, respectively), as compared to March 2020 early in the pandemic, though neither result is statistically significant [Sharma and Khokhar 2021; Disaster Medicine and Public Health Preparedness].

Egger et al. 2021

In a sample of rural households in Kenya, Eggers et al. find an increase in violence against both women and children during the crisis period (of 4% and 13%, respectively), as compared to March 2020 early in the pandemic, though neither result is statistically significant [Egger et al. 2021; Science Advances].

Venter et al. 2020

Reviewing the patient register at a hospital emergency department in Johannesburg, South Africa from February to June 2019 and the same period in 2020, Venter et al. note that cases of trauma from interpersonal violence decreased by 25% year over year. However, results are not statistically significant, nor are they sex-disaggregated or disaggregated by type of violence [Venter et al. 2020; South African Medical Journal].

Papers exploring the experience, risk factors, and prevention of VAW/C during COVID-19

An additional eleven papers assess risk factors, report on experiences of violence, or examine prevention programming during the pandemic. Of studies documenting significant risk factors for increased violence, several of note were being married, being unemployed (for either the victim or perpetrator), having lost household income due to the pandemic, and the perpetrator’s substance abuse tendencies. Of particular note are studies coming from across a range of country contexts that point to heightened economic vulnerability—whether in the form of unemployment, reduced household income, or food insecurity—as tied to a greater risk of violence. As dynamics and lived experiences broaden beyond containment to include widespread economic crises, effects on VAW/C will also continue to evolve, likely in ways that outlast the direct health effects of COVID-19 and associated containment measures.

Few studies point to potential protective factors, but evidence from India and Ethiopia suggests that higher education (for both victim and perpetrator) decreases the risk of violence. Levels of IPV were twice as high for illiterate women in Ethiopia as they were for women who had completed secondary school, and in India rates of domestic violence were significantly lower if either the husband or wife had an advanced degree. Employment status may also be a protective factor, with studies in Jordan and Iran showing significantly lower rates of violence for women who are employed. This mirrors the risk factor of losing employment or income during the pandemic being associated with increased violence.

Two studies focused on violence against healthcare workers (in Iran and China), and two studies focused on pregnant women specifically (in Iran and Ethiopia), all documenting high rates of violence during the pandemic. Only one study included an evaluation of a prevention or harm-mitigation program, a youth empowerment program in Bolivia, finding a reduction in violence experienced by girls of nearly ten percentage points (or 46% relative to the control group), seven months after its completion.

Group B. VAW/C experiences, risk factors, and prevention during COVID-19

Authors Location Data Sample size Key findings
1. Ghanbari et al. 2020 Rasht, Iran Survey data (self-administered) 112 emergency room nurses who work directly with COVID-19 patients Prevalence of verbal abuse of nurses was 62.5% over the first 6 months of the pandemic and usually perpetrated by the patients or their families. Prevalence of physical violence was 17.8%.
2. Gulesci et al. 2021 Bolivia Survey data (in-person; telephone) 511 adolescents Girl participants in a youth empowerment program in Bolivia are 9.6 percentage points less likely to report experiencing violence relative to girls in a control group, while boys saw no significant change.
3. Haddad et al. 2020 Lebanon Survey data (online) 369 women Women subject to psychological violence during the COVID-19 lockdown have a lower, but not significantly lower, probability of pregnancy and a higher probability of unwanted pregnancy.
4. Hajj et al. 2021 Lebanon Survey data (online) 502 adults Violence at home was significantly associated with increased stress and insomnia, and weakly associated with anxiety and well-being, and no significant association with post-traumatic stress symptoms.
5. Krishnakumar & Verma 2021 India Newspaper articles 59 newspaper articles Prominent themes that emerged from analyzing newspaper reports of domestic violence incidents were withdrawal from alcohol and unemployment for the perpetrator.
6. Mahapatro et al. 2021 Alwar, India Qualitative data (phone) 36 women survivors of domestic violence Women survivors found it much more difficult to access services and social support networks to deal with domestic abuse during the lockdown period.
7. Naghizadeh et al. 2021 Tabriz, Iran Survey data (in-person) 250 pregnant women During COVID-19, 35.2% of pregnant women surveyed had experienced domestic violence, including 32.8% who had experienced emotional violence, 12.4% experienced sexual violence, and 4.8% experienced physical violence. Experiences of domestic violence were correlated with reduced spouse income during the pandemic.
8. Rockowitz et al. 2020 Kenya Survivor intake forms (from violence services) 317 adult survivors, 224 child survivors During the pandemic, children were more likely to be attacked during the day, in private, by a single perpetrator who was most often a family member or neighbor. Adults were equally likely to be attacked by strangers and persons known to them, often in public.
9. Tadesse et al. 2020 Dessie, Ethiopia Survey data 589 married or cohabited women Approximately 22% of respondents experienced at least one form of IPV (physical, psychological, and sexual) during lockdown. The most significant determinants of having experienced violence were being illiterate or having an illiterate husband, having a substance user husband, and community tolerance of violence.
10. Teshome et al. 2020 Addis Ababa, Ethiopia Survey data 464 pregnant women Prevalence rate of IPV of pregnant women sampled was 7.1%, and among them 72% reported emotional violence, 49% reported sexual violence, and 30% reported physical violence. One significant predictor of IPV was having a husband who chewed Khat and drank alcohol.
11. Wang et al. 2020 China Survey data (online) 1,063 healthcare workers Rates of medical workplace violence were 20.4% during the COVID-19 outbreak, and those who had experienced workplace violence were more likely to have elevated mental health problems.

More details on papers exploring experiences of VAW/C during COVID-19

Ghanbari et al. 2020

Using survey data from 112 emergency department nurses in Rasht, Iran, Ghanbari et al. find that 62.5% of nurses experienced verbal violence and 17.8% experienced physical violence at work during the first six months of the pandemic, mostly occurring during the evening and night shifts. Many nurses did not report abuse because they felt it would not be useful [Ghanbari et al. 2020; Working paper].

Gulesci et al. 2021

Girl participants aged 15 to 18 in a youth empowerment program in Bolivia are 9.6 percentage points (46%) less likely to report experiencing violence relative to girls in a control group (driven by psychological and sexual violence), while boys saw no significant change, seven months after the intervention ended. The program included modules on personal empowerment, sexual and reproductive health, economic empowerment, technical skills training, and work insertion/business development and the authors hypothesize the decrease in experiences of violence may be due to an increase in earnings and bargaining power within the home [Gulesci et al. 2021; World Bank].

Haddad et al. 2020

In a study of 369 Lebanese women on the drivers of pregnancy outcomes during the COVID-19 pandemic, Haddad et al. find that experiences of psychological violence are negatively associated with women becoming pregnant during containment measures, but positively associated with unwanted pregnancies, though neither result was statistically significant [Haddad et al. 2020; Working paper].

Hajj et al. 2021

Using survey data from 502 adults in Lebanon, Hajj et al. find that more men than women reported experiencing violence (8.4% vs. 3.8%), and that those who had experienced violence were more likely to experience increased stress and insomnia (note the latter results were not sex-disaggregated) [Hajj et al. 2021; Working paper].

Krishnakumar & Verma 2021

Using 59 newspaper articles from across India, Krishnakumar and Verma identify alcohol withdrawal and unemployment to be main drivers of domestic violence during the lockdown period. They also note that lockdowns made women particularly vulnerable by isolating them from friends or family who might otherwise intervene or provide support [Krishnakumar & Verma 2021; Asian Journal of Criminology].

Mahapatro et al. 2021

Through 36 qualitative interviews, Mahapatro et al. find that access to services for domestic abuse survivors was severely limited during the lockdown period in India and that only emergency services could be provided by phone. Additionally, lockdowns prohibited women from accessing social networks, which they would normally rely on for support to cope with violence [Mahapatro et al. 2021; Journal of Family Issues].

Naghizadeh et al. 2021

Through in-person surveys at an obstetrics clinic in Tabriz, Iran, Neghizadeh et al. find high prevalence rates of domestic violence among pregnant women (35.2%) during the COVID-19 pandemic. Broken down further, 32.8% of women surveyed had experienced emotional violence, 12.4% had experienced sexual violence, and 4.8% had experienced physical violence. The mean score on a scale of mental health was significantly lower for those women who had experienced violence during COVID-19 than those who had not. Additionally, reduced spousal income was positively associated with experiences of domestic violence during the pandemic [Naghizadeh et al. 2021; BMC Pregnancy and Childbirth].

Rockowitz et al. 2020

Using survivor intake forms from violence-related services, Rockowitz et al. find that children are more likely to be attacked during the daytime in private by a single family member or neighbor. Adults are equally likely to be attacked by a stranger or a person known to them and in public. [Rockowitz et al. 2020; PsyArXiv Pre-print].

Tadesse et al. 2020

In a random sample of married or cohabited women in Ethiopia, Tadesse et al. find an IPV prevalence rate of 22.4%, with the most significant determinants of having experienced violence were being illiterate or having an illiterate husband, having a substance user husband, and community tolerance of violence. Broken down further, 20% of respondents had experienced psychological violence, 13.8% had experienced sexual violence, and 11% had experienced physical violence during lockdown. [Tadesse et al. 2020; Journal of Interpersonal Violence].

Teshome et al. 2020

Through an in-person survey of pregnant women at a prenatal clinic in Addis Ababa, Ethiopia, Teshome et al. find that 7.1% of their sample had experienced IPV during the pandemic, with 72% of those reporting emotional violence, 49% reporting sexual violence, and 30% reporting physical violence. Less than 2% of participants were screened for IPV at the clinic. Spousal consumption of Khat or alcohol was associated with IPV incidence [Teshome et al. 2020; International Journal of Gynecology and Obstetrics].

Oguntayo et al. 2020

Using data collected via online survey in Nigeria among 356 adults in Lagos, Nigeria, Oguntayo and colleagues show correlations between IPV and both personality traits (neuroticism), as well as socio-contextual factors (poorer living conditions and less stable employment), though no differences by gender [Oguntayo et al. 2020; International Journal of Behavioral Sciences].

Wang et al. 2020

Using an online survey of healthcare workers in China, Wang et al. find that 20.4% of those surveyed had experienced workplace violence during the COVID-19 outbreak. Female health care workers were significantly less likely to report having experienced workplace violence (than male healthcare workers), though both were equally likely to report mental health problems after experiencing workplace violence, using a propensity score matching approach [Wang et al. 2020; Risk Management and Health Care Policy].

Evidence gaps one year on: A need to focus on prevention and mitigation

While the number of papers examining trends and risk factors for VAW/C during the pandemic continues to grow, after ‘rounding up’ 100 papers, the major gap continues to be evidence on prevention and mitigation. As many countries shift their focus towards recovery, a new set of questions and challenges are emerging. What types of policies and programs are effective during the pandemic and the recovery to mitigate and prevent diverse forms of VAW/C—including for specific populations (e.g., adolescent girls; healthcare workers)? How cost effective are these efforts, given programming tradeoffs? Are there important lessons for longer-term programming efforts that go beyond current experiences, to inform how to mitigate intergenerational and long-term effects? This necessitates a focus not only on more complex data collection and analysis building on new or existing impact evaluations, but also offers a role for rigorous qualitative work. Only two of the 26 papers reviewed in this round up collected and analyzed qualitative data, a methodology which can help narrate the lived experiences of women and children reached by interventions.

What do we know so far from intervention studies? A total of five papers have assessed the effectiveness of response efforts thus far: a helpline campaign in Italy, stimulus payments and differential firearm policies in the United States, alcohol consumption policies in Mexico, and a youth empowerment program in Bolivia. The findings are mixed. A ban on alcohol was found to have no impact on the number of calls to seek domestic violence services during lockdown, while the awareness campaign in Italy was associated with an increase in calls to a domestic violence helpline. Both of these studies point to changes in help-seeking rather than changes in the prevalence in VAW/C experience itself. In the United States, daily domestic violence calls decreased significantly after stimulus payments were made, but remained high in areas with higher concentrations of Hispanics and noncitizens (who may face higher barriers to accessing the welfare system). More stringent gun laws also seemed to mitigate the increase in domestic violence seen in the United States, however this finding may not translate to other settings with lower numbers of firearms or existing firearm restrictions. The Bolivia study is the first experimental evaluation of a targeted intervention during COVID-19, and with promising impacts for adolescent girls specifically.

While results are likely to differ across settings, future research efforts and funding should focus on response—across a promising range of programming, while prioritizing the safety of study participants—rather than continue to question if levels and risk factors have increased. Studies examining risk and protective factors provide a starting point that can help researchers prioritize interventions to evaluate going forward. Given the number of studies pointing to economic insecurity’s ties to risk of violence, researchers can consider prioritizing the evaluation of policies and programs aimed at supplementing household income and food consumption, including through the provision of cash and in-kind transfers and employment opportunities. Additionally, with many studies identifying alcohol consumption, substance use and mental health as linked to experiences of violence, more work should be done to determine the best policies to mitigate these risks.

The authors thank David Evans for helpful comments.

Rights & Permissions

You may use and disseminate CGD’s publications under these conditions.