December 2020 What Factors Exacerbate and Mitigate the Risk of Gender-Based Violence EAST ASIA AND PACIFIC GENDER INNOVATION LAB During COVID-19? The East Asia and Pacific Gender Insights From a Phone Survey in Indonesia Innovation Lab (EAPGIL) carries out impact evaluations and KEY FINDINGS inferential research • 83% of respondents report increase in Intimate Partner Violence in their to generate evidence communities due to COVID-19 on what works in • Household food insecurity is among the strongest predictors of exposure to gender- closing gender gaps based violence in assets, economic • Women’s access to jobs protects them from increase in exposure to gender-based opportunities, and violence due to COVID-19 agency, and how closing these gaps can help achieve CONTEXT other development 1 in 3 Indonesian women have experienced Gender-Based Violence (GBV) in their lifetime.1 outcomes. Ultimately, The COVID-19 pandemic may further exacerbate the risks of GBV. First, additional stress due EAPGIL seeks to to health risks and economic uncertainty is likely to trigger conflict within family. Second, more increase the welfare time spent in the same physical space with potential perpetrators due to lockdowns may also of women and men increase the likelihood of abuse. Indeed, in a study of 15 countries, UN Women found that in East Asia and the calls to GBV hotlines surged in 12 countries, from 40% in Malaysia to 400% in Tunisia.2 Pacific by promoting Even before COVID-19, there was no regular systematic data collection on GBV in Indonesia the uptake of effective aside from the one-off study by Statistics Indonesia in 2016,3 which makes it difficult to policies and programs understand how the COVID-19 pandemic might have increased the incidence of GBV. identified based on 1  Statistics Indonesia (2016), 2016 Indonesian National Women’s Life Experience Survey (Survei Pengalaman Hidup evidence. Perempuan Nasional). The survey was joint work with the Ministry of Women Empowerment and Child Protection and UNFPA. UN Women (2020), Impact of COVID-19 on violence against women and girls and service provision: UN Women rapid 2  assessment and findings. In many other countries, data on GBV often come from Demographic and Health Surveys (DHS). DHS in Indonesia 3  has never collected data on GBV. The most recent study on GBV in Indonesia was implemented jointly by Statistics Indonesia, the Ministry of Women Empowerment and Child Protection and UNFPA in 2016, 2016 Indonesian National Women’s Life Experience Survey (Survei Pengalaman Hidup Perempuan Nasional). Anecdotal evidence from one hotline in Jakarta suggested of Indonesia’s Desmigratif program, which set up an increase in GBV, with the hotline receiving 110 calls migration information centers in 400 villages across reporting domestic violence between March and June 2020 – the country over a three year period (2017-2019). The amounting to 50% of calls received in the entirety of 2019.4 program specifically targeted villages with high shares of We collected data on exposure to GBV through a phone international migrant workers. In Indonesia, international survey to understand the factors that pose the greatest risk migrant workers are largely low-skilled from rural areas— and policy interventions that may effectively protect women. with domestic, farm, construction, and factory workers In-person data collection was not possible due to health comprising almost 80 percent of the migrant workers.9 concerns associated with the COVID-19 pandemic. In order The phone survey is certainly not nationally to not jeopardize the safety of the respondents through representative and the findings should be interpreted backlash from perpetrators living in the same households, we did not ask questions about violence directly. Rather, within this sub-sample of the population. 88 percent based on consultations with GBV experts, we developed a of individuals in the phone survey lived in rural series of proxy questions, which allowed us to infer the likely areas, compared with just 44 percent in the national exposure to violence. 5, 6 population.10 The phone survey was administered to the same individuals who were interviewed in the DATA previous 2018 survey. Thus, we could use a rich set We administered interviews to 866 women in a phone survey of pre-COVID-19 characteristics from the 2018 survey across 6 provinces in Indonesia.7 We also collected data on in our analysis. The phone survey data was collected their households. We were able to reach our respondents 8 in late July – early September 2020, during a period using phone numbers collected in August-November of relaxed social restrictions, after stricter lockdowns, 2018 for an ongoing impact evaluation of the Government imposed between April – July 2020, had ended.11 Figure 1: COVID-19 Exacerbated Women’s Perceived Risks of Violence in Indonesia Experienced injury* 17 4 Unsafe at home 43 Unsafe in 16 commmunity 46 Frequent conflict 8 at home 18 0 10 20 30 40 50 60 70 80 90 100 Share of Respondents (%) Experience in the past 6 months (Mar-Aug 2020) Worsened due to COVID 4 https://theconversation.com/angka-kdrt-di-indonesia-meningkat-sejak-pandemi-covid-19-penyebab-dan-cara-mengatasinya-144001 5 We are grateful to Amber Peterman (University of North Carolina at Chapel Hill) and Diana Arango (the World Bank) for insightful comments on the instrument. The team has also followed protocols based on the WHO recommendations to ensure safety of the respondents, such as training of enumerators to inform 6  respondents on certain keywords that can be used to stop the survey in case third parties start listening into the conversation. 7 We collected data in West Java, Central Java, Yogyakarta, East Java, East Nusa Tenggara, South Sulawesi, 88% of households reached are in rural areas. 8 Either from the woman, or from another sufficiently knowledgeable household member. 9 World Bank (2017), Indonesia’s Global Workers: Juggling Opportunities and Risks. World Bank World Development Indicators. https://databank.worldbank.org/source/world-development-indicators 10  Lockdowns were not imposed nationally, but by individual local governments at province and district levels. Exact dates for initial lockdowns vary but for our 11  surveyed districts are within the April-July timeframe. Figure 2: Vignettes Similarly Showed that Risks of Violence Were Becoming More Common in the Community Violence in the 43 community against intimate partner 83 Violence in the 50 community against children 68 Harrassment in 28 the community 65 0 10 20 30 40 50 60 70 80 90 100 Share of Respondents (%) Common or very common in the community Worsened due to COVID EXPOSURE TO VIOLENCE and BUDI have been married for several years and have two children. BUDI works in a repair shop, but lately We attempt to capture exposure to violence during the the business has been bad, and they are worried about 6 months preceding data collection: from the onset of money. Sometimes when BUDI gets stressed, he takes COVID-19 in March 2020 until the interview. We gauge out his anger by yelling at IRMA, and sometimes he hits exposure to violence by asking 4 questions: her. IRMA feels hurt and wants him to stop but does not 1. In the last 6 months, have you been injured in any way? know what to do.” After the vignette, the respondent was For example, have you had cuts, bruises, aches, burns, asked: How common do you think it is for couples in your sprains, dislocations, broken bones, or any other wound community to experience such a story? We also included that limited your functioning? vignettes structured to capture violence against children 2. In the last 6 months, did you feel safe in your home? and harassment.12 3. In the last 6 months, did you feel safe in your For each question aimed to capture exposure to community? violence, except on experience of injury, we followed up with a question asking whether “COVID-19 has made the 4. When people live together in the same household, they situation worse, better, or left it the same.” This follow- usually share both good and bad moments. And it is up question was designed to capture respondent’s normal for people who live together to have arguments. subjective perception of changes in violence due to the How often in the last six months would you say that COVID-19 pandemic. people in your household have argued or have had some We found that 17% of women in our sample experienced sort of conflict among themselves? (The answer options injury, 4% felt unsafe at home, 16% felt unsafe in the are: Never, once or twice, weekly, daily, don’t know) community, and 8% experienced conflicts at least once To increase the likelihood of reporting, we also administered a week during 6 months prior to the interview (between several vignettes, where we described a hypothetical March and August 2020, Figure 1)13. Notably, responses to situation, involving exposure to violence, and asked the vignettes suggest higher exposure to violence in their how frequently such situations were likely to occur in a community: 43%, 50% and 27% of the respondents agree respondent’s community. For example, the vignette aimed that IPV, violence against children and harassment are to capture intimate partner violence (IPV) reads as: “IRMA common or very common in their communities (Figure 2). Exact text of the vignettes is available at https://documents.worldbank.org/en/publication/documents-reports/documentdetail/950601606987399330/can- 12  we-capture-exposure-to-gender-based-violence-gbv-through-phone-surveys-during-a-pandemic Ideally, we would compare these estimates with the data on GBV from a different source. However, unfortunately, the Demographic and Health Surveys in 13  Indonesia do not include a GBV module. The EAPGIL team is exploring the possibilities of comparing the rates of GBV captured using this method to GBV rates captured through DHS surveys in other countries, such as Lao PDR. Our data also capture the perception that violence increased Economic stress increases the likelihood of due to COVID-19. Figure 1 shows that 43% and 46% of violence respondents feel less safe at home and outside of home, Food insecurity experienced by the household and respectively; 18% reported more frequent arguments due the number of household members are among the to COVID-19. 83%, 68% and 65% shared perception that most important predictors of exposure to GBV. COVID-19 increased likelihood of IPV, violence against These results are aligned with existing theoretical children and harassment in their communities (Figure 2). frameworks, which posit that economic insecurity is an important determinant of domestic violence WHAT FACTORS INCREASE AND (Ellsberg et al., 2015; Buller et al., 2018). Food MITIGATE THE LIKELIHOOD OF insecurity increases such stress. Higher number of EXPOSURE TO GBV SINCE THE ONSET OF household members, in most cases, implies more THE COVID-19 PANDEMIC? children and elderly, likely also augmenting stress. Understanding what factors exacerbate GBV and what Having a job is among the strongest factors lower its likelihood is important for design of relief protective factors from increase in violence and recovery policies. With such insights, policy makers due to COVID-19 may adjust their response to also lower the risks of GBV, in Having a second job is the strongest protective addition to other objectives of immediate relief, protecting factor from increase in violence due to the COVID-19 human capital and economic recovery. pandemic. There are two theoretical explanations To understand which factors are likely to trigger or mitigate for this finding. On the one hand, additional income GBV, we leveraged our rich datasets, collected through may mitigate economic stress. On the other hand, in-person interviews in 2018 and phone survey interviews theories of intra-household bargaining predict that in 2020. The datasets include information on employment, women’s independent income is likely to reduce GBV non-agricultural enterprises, remittances, food security, (Manser and Brown, 1980). As a woman’s potential social assistance, knowledge of COVID-19, domestic work, options outside of marriage improve due to her and health symptoms. We constructed an index of exposure economic empowerment, her situation within marriage to GBV and an index of increased intensity of GBV due to is expected to get better, too. More economically COVID-19, based on all diverse proxy and vignette variables empowered women have an option to leave an included in the survey.14 abusive relationship, which increases their bargaining power within the relationship, thus, decreasing We then used a machine learning algorithm to sift through violence. 156 variables from our rich datasets and identify which of these are important predictors of exposure to GBV and Women in highly populated urban areas may increased intensity of GBV due to COVID-19. The machine be at a lower risk learning algorithm, however, does not detect the direction of We find that women in districts with higher COVID-19 the effect, only the strength of association.15 To understand risk level17 were less likely to report exposure to which of the top predictors work as a protective or a risk violence. These are likely to be urban districts with factor, and to assess relative magnitudes, we carried out higher population density (Olivia, Gibson, Nasrudin, a stepwise linear regression analysis using the strongest 2020) – two characteristics that our dataset does not 20 predictors. The stepwise regression analysis further allow us to include directly into our analysis. Such dropped relatively weaker predictors and identified which districts are likely to offer women better access to had statistically significant relationship with risk of GBV and economic resources, institutional support, and more increase in this risk due to COVID-19. Several important gender equitable social norms, which have been patterns emerged. 16 shown to lower the risk of GBV (Mcllwaine, 2013). 14 We follow the methodology used by Kling, Katz, and Liebmann (2007), where the indices are constructed as equally weighted mean of z-scores of the components. The GBV questions administered consist of 2 types of questions, those proxying exposure, and those proxying change in exposure. We construct separate index for each group of questions, with each component of the indices oriented in such a way that higher values indicate worse/ worsening GBV. We used all questions described in section 3, except for the question on violence against children. 15 We use Random Forests algorithm, which allows us to detect whether food insecurity is associated with GBV more strongly than, for example, household size. However, the Random Forests algorithm does not reveal whether increase in food insecurity is associated with increase or decrease in GBV. 16 Technical details available upon request from England Rhys Can at englandrhys@worldbank.org. 17 These data were retrieved through data scraping from: https://covid19.go.id/peta-risiko Figure 3: Percentage of Women Who Believe Beating by Husband is Justified for Going Out Without Permission (by income quintile) 26 24 22 20 18 16 14 12 10 8 6 4 2 0 Quintile 1 Quintile 2 Quintile 3 Quintile 4 Quintile 5 Source: World Bank Gender Data Portal / Data for Indonesia, 2017 The perils of the middle: women with lower and empowering women—for example, financially—which higher education and age are at lower risk than may protect them from GBV. their counterparts in the middle of education and Similarly, the relationship between women’s age and risk age distributions of worsening of violence due to COVID-19 also follows a Our results suggest that women with 11 years of similar inverted U pattern. education are at a higher risk of increase in violence due to COVID-19, compared to women with less or more years of education. Specifically, likelihood of increase in INSIGHTS FOR POLICY violence due to COVID-19 goes up as the number of years We collected data on proxies of exposure to GBV of education increases. However, once women reach between March and August 2020 and the perceived 11 years of schooling, every additional year works as increase in the intensity of GBV due to the COVID-19 protective factor. pandemic through phone interviews. We subsequently carried out exploratory analysis of a large data set Several factors contribute to this inverted U-shape with the objective of furthering the understanding of relationship. First, women may define GBV differently what factors may protect women or increase the risk depending on their education level. Despite being exposed of exposure to GBV during the pandemic. Our results to violence, lower educated women may not necessarily suggest few implications for policy: view the situation as violent or abnormal. For example, the likelihood to perceive wife-being as justifiable decreases 1. It is critical to expand and continue the provision of as household income increases in Indonesia (Figure 3). social protection measures during the pandemic. The Thus, lower educated women may under-report their Government of Indonesia rolled out various social exposure to GBV. As education increases, women’s assistance measures to mitigate socio-economic risks perception of violence may change. At the same time, the of the COVID-19 pandemic for the most vulnerable very fact of getting more education may be perceived as families, such as staple foods packages, cash “breaking the norms” and the status quo that women, for transfers, electric bill subsidies, and wage subsidies. instance, should primarily be caretakers, not needing too These programs not only provide immediate financial much education; putting them at a higher risk of GBV. But support to all family members but may also protect after a certain level, additional years of education start women from GBV. Food insecurity experienced by the household is among the strongest predictors of exposure to GBV. Bold measures ACKNOWLEDGMENTS taken by the Government of Indonesia could help mitigate some concerns related to food and economic insecurity induced by the COVID-19 pandemic, which in turn This brief is a product of could reduce conflicts within the family and lower the likelihood of GBV. collaboration between EAPGIL, Gender CCSA, Poverty GP 2. Policies fostering women’s economic empowerment should be implemented both and DECDG. It was prepared during the pandemic and as recovery measures. Aside from boosting economic by Daniel Halim, England Rhys growth, protecting gains in women’s economic empowerment also protects them Can and Elizaveta Perova. from GBV. Women who had access to more jobs during the COVID-19 pandemic were less likely to perceive an increase in the exposure to GBV due to COVID-19. We gratefully acknowledge As women are shouldering a greater share of childcare responsibility during the funding from the Umbrella pandemic, it is important to create policy responses that will protect women’s gains Facility for Gender Equality in the labor market from the blow of the pandemic. For example, the availability of (UFGE) to carry out this low-cost public preschools in Indonesia had been shown to increase women’s work work. EAPGIL is supported participation.18 Improving access to affordable and quality childcare services, closer by UFGE in partnership with to homes or workplaces with extended hours and quality assurance system, could the Australian Department help increase the demand for childcare services, allowing women to work and as of Foreign Affairs and evidence shows – reducing the risk of GBV. In addition, the Government of Indonesia Trade. UFGE has received may consider extending paid maternity and paternity leave benefits.19 Maternity generous contributions from leave benefits could help women transition better to motherhood without necessarily Australia, Bill & Melinda exiting the workforce, while non-existent or minimal paternity leave benefits risks Gates Foundation, Canada, discouraging firms from hiring more female employees. Denmark, Finland, Germany, Iceland, Latvia, Netherlands, REFERENCES: Norway, Spain, Sweden, Katz, Lawrence, Jeffrey R Kling, and Jeffrey B Liebman. 2007. Experimental Analysis of Switzerland, United Kingdom, Neighborhood Effects. Econometrica, Volume 75, Issue 1. and the United States. Manser, Marylin and Murray Brown. 1980. Marriage and Household Decision-Making: A Bargaining Analysis. International Economic Review. Vol. 21, No. 1. McIlwaine, Cathy. 2013. Urbanization and gender-based violence: exploring the paradoxes in the global South. Environment and Urbanization. Volume 25, Issue 1. Olivia, Susan, John Gibson and Rus’an Nasrudin. 2020. Indonesia in the time of COVID-19. Bulletin of Indonesian Economic Studies, Vol. 56, No. 2, 2020: 143–174 FOR MORE INFORMATION Elizaveta Perova eperova@worldbank.org Daniel Halim dzhalim@worldbank.org Halim, Johnson, Perova (2019). Preschool Availability and Female Labor Force Participation: Evidence from 18  http://www.worldbank.org/eapgil Indonesia. In 2019, the laws in Indonesia mandated 90 days of paid maternity leave and 2 days of paternity leave 19  (Women, Business, and the Law 2020). These are lower than the global average of 109 days and 8 days of maternity and paternity leave benefits, respectively.