Effet de la violence contre les femmes sur les victimes et leurs enfants : Données d'Amérique centrale, de la République dominicaine et d'Haïti

Effect of Violence against Women on Victims and their Children Boaz Anglade Julia Escobar TECHNICAL NOTE N o IDB-TN-2139 March 2021 Country Department Central America, Haiti, Mexico, Panama and the Dominican Republic Evidence from Central America, the Dominican Republic, and Haiti Inter - American Development Bank Effect of Violence against Women on Victims and their Children Boaz Anglade Julia Escobar Inter-American Development Bank Country Department Central America, Haiti, Mexico, Panama and the Dominican Republic March 2021 Evidence from Central America, the Dominican Republic, and Haiti Cataloging - in - Publication data provided by the Inter-American Development Bank Felipe Herrera Library Anglade, Boaz. Effect of violence against women on victims and their children: evidence from Central America, the Dominican Republic, and Haiti / Boaz Anglade, Julia Escobar. p. cm. — (IDB Technical Note; 2139) Includes bibliographic references. 1. Women-Violence against-Central America. 2. Women-Violence against-Dominican Republic. 3. Women-Violence against-Haiti. 4. Women-Crimes against-Central America. 5. Women-Crimes against-Dominican Republic. 6. Women-Crimes against- Haiti. 7. Intimate partner violence-Central America. 8. Intimate partner violence- Dominican Republic. 9. Intimate partner violence-Haiti. I. Escobar, Julia. II. Inter- American Development Bank. Country Department Central America, Haiti, Mexico, Panama and the Dominican Republic. III. Title. IV. Series. IDB-TN-2139 http://www.iadb.org Copyright © 2021 Inter-American Development Bank. This work is licensed under a Creative Commons IGO 3.0 Attribution- NonCommercial-NoDerivatives (CC-IGO BY-NC-ND 3.0 IGO) license ( http://creativecommons.org/licenses/by-nc- nd/3.0/igo/legalcode ) and may be reproduced with attribution to the IDB and for any non-commercial purpose. No derivative work is allowed. Any dispute related to the use of the works of the IDB that cannot be settled amicably shall be submitted to arbitration pursuant to the UNCITRAL rules. The use of the IDB’s name for any purpose other than for attribution, and the use of IDB’s logo shall be subject to a separate written license agreement between the IDB and the user and is not authorized as part of this CC-IGO license. Any dispute related to the use of the works of the IDB that cannot be settled amicably shall be submitted to arbitration pursuant to the UNCITRAL rules. The use of the IDB’s name for any purpose other than for attribution, and the use of IDB’s logo shall be subject to a separate written license agreement between the IDB and the user and is not authorized as part of this CC-IGO license. Note that link provided above includes additional terms and conditions of the license. The opinions expressed in this publication are those of the authors and do not necessarily reflect the views of the Inter -American Development Bank, its Board of Directors, or the countries they represent. 1 Abstract 1 This paper presents a systematic overview of the evidence of violence against women in the Central America, Mexico, Panama, Haiti, and Dominican Republic region and examines its impact on the well - being of women and their children. Population - based surveys show that violence against women remains a widespread issue in the region. The proportion of women who have experienced physical or sexual violence at least once in their lifetime varies between 13% and 53%; Panama has the lowest rate while Mexico and El Salvador have the highest. The percentage of women who have experienced violence within private spheres ranges between 17% and 24%. Also, homicidal violence targeting women remains a major problem in the region. Using a novel propensity score reweighting t echnique, we assess the impact of violence on a series of outcome variables related to a woman’s health and socioeconomic condition. We find evidence that violence against women negatively affects victims’ reproductive and physical health as well as their fertility preferences. We also find evidence that violence against mothers has an adverse effect on children’s advancement in school and overall health. Classification JEL: C50, I10, I30, J16 Keywords: intimate partner violence, gender - based violence, v iolence against women, femicide, covariate balancing, propensity score reweighting. 1 The authors are grateful to Laurence Telson, Nidia Hidalgo, and other colleagues at the IDB’s Gender and Diversity (GDI) Division for their constructive comments on previous versions of this paper. 2 Introduction Vio lence against women (VAW) is not just a flagrant violation of human rights but also a serious issue affecting both public health and the economic development of countries. It is defined as any act of violence based on gender that results in physical, sexua l, or psychological harm or suffering to women. In the literature, the terms violence against women and gender - based violence (GBV) are often used interchangeably as evidence from around the world points out that violence based on gender is an issue that affects women disproportionately. Whereas men are more likely to experience violence due to conflict and crime, women are more likely to experience sexual violence and violence from individuals in their close circle (Bott et al., 2012). VAW is therefore deeply rooted in inequality and is a consequence of the unequal distribution of power between the sexes. VAW is an issue prevalent throughout the world, in both low - and high - income countries, with serious consequences for women’s health and overall well - being. One of the most common forms of VAW is intimate partner violence (IPV). It refers to any abusive behavior by a cur rent or previous intimate partner that causes physical, sexual, or psychological harm. While the VAW measure paints a full picture of the extent of violence experienced by women in everyday life (inside and outside of the home), the IPV measure focuses on occurrences inside the home and in intimate circles. A multi - country study published in 2005 by the World Health Organization (WHO), using a standard instrument, shows IPV prevalence rates ranging from 15 to as high as 71%. A subsequent 2013 report indicat es that one in three women globally have experienced violence at least once in their lifetime at the hand of an intimate partner. Violence against women is a source of concern in many developing regions globally, particularly in Latin America. The 2013 WHO report classifies the Americas (LAC) as the next highest region in terms of IPV prevalence (30%), following the African, Eastern Mediterranean, and South - East Asian regions. Femicide, defined as homicidal violence targeting women, is the most extreme type of violence against women and a major concern in the LAC region. The Central American countries of Honduras and El Salvador, in particular, have the highest rates of femicide in the region (Figure A4). In recent years, there has been increasing interest f rom academic researchers and policymakers alike in documenting and assessing the impact of gender - based violence on individuals' well - being, particularly women. Several studies on VAW have been conducted in developed and developing contexts, and important strides have been made over the years in the quantity and rigor of data collected. The number of countries with national VAW estimates has grown considerably over the years. From 1995 to 2014, 102 countries have conducted at least one VAW survey, either separately or as a module to a larger household survey (United Nations, 2015). Some important methodological advances have also been made in addressing various well - known difficulties attached to collecting data in developing countries, particularly issues related to ethics in data collection and sensitivity of information. One of the remaining challenges to researchers is the difficulty in comparing estimates across geography and time. Given that data collection efforts rely on different methodologies, estimates are sometimes presented in diverse ways, and researchers and policymakers often lack access to comparable estimates even when reliable data 3 exist. For example, different survey methodologies use different definitions of partnership status and collect information based on different age groups, different forms of violence, different partnership status (current partner or most recent partnership) , and different time frame of perpetration (at least once in a lifetime or in the 12 months prior to data collection). Why should we be interested in assessing the economic impact of VAW in the first place? VAW not only represents a threat to public health and safety but can also have serious consequences on economic variables, thus impeding progress towards economic development. VAW can have significant costs to an economy in terms of loss of income, decreased productivity, increased expenditure on service s, and decrease in human capital formation, among other things. On a microeconomic level, its consequences can be wide - ranging and long - lasting on victims and their children's well - being. Studies have shown that violence leads to serious adverse outcomes i mpacting women’s reproductive, mental, and physical health (Day et al., 2005). It has serious public health ramifications as well, given that it increases the risk of morbidity and mortality among women (Heise et al., 2002). Violence also has intergenerati onal effects. There is overwhelming evidence that children who witness violence are at an increased risk of becoming victims or even aggressors themselves; they are also more prone to perform poorly in school, thus limiting their future labor market possibilities (Morrison & Biehl, 1999). Research on the consequences of violence is important and should be encouraged. Existing efforts to prevent and end VAW will be limited without an understanding of the impact o f such violence. Findings could allow policymakers and practitioners to provide more comprehensive responses to gender - based violence issues and could also help guide future initiatives. This paper provides an overview of the evidence on violence against w omen in the Central America, Mexico, Panama, the Dominican Republic, and Haiti (henceforth CID) region. As such, we conduct a systematic review of the prevalence of VAW across countries in the region and over time. We consider two forms of violence: physical and sexual (both over a lifetime and in the past 12 months) and discuss two VAW measurements: a broader VAW prevalence measure and a narrower IPV measure. We also take a closer look at the prevalence of femicide, which is also a serious issue affecting the region. Further, we assess the impact of VAW on various outcomes related to women's socioeconomic and health conditions and their children’s development. Although this chapter covers countries in the CID region, the empirical analysis i s restricted to four countries (Haiti, Dominican Republic, Guatemala, Honduras) for which comparable data are readily available. 4 VAW in the Context of COVID - 19 On March 11, 2020, the World Health Organization declared COVID - 19, a viral respiratory infection, a global pandemic. Governments around the world imposed mandatory lockdowns and shelter - in - place measures to mitigate the propagation of the viral disease. The pandemic has brought to the fore of public consciousness the bitter realization that home is not always a safe space for many women, and restrictive or quarantine measures can come with increased risks for women and girls. With the social distancing measures and mobility restrictions (Left chart) implemented in Central America and Mexico came a significant increase in calls to domestic violence helplines (right chart). Mexico experienced a 32% jump in domestic violenc e calls during t he lockdown. Central American countries such as Guatemala, El Salvador, and Costa Rica also reported a 9% to 30% increase in calls to police helplines. In Brazil, reports of domestic violence increased by 40% to 50%; calls to domestic violence hotlines als o increased in Argentina (40%) (Campbell, 2020). Recent empirical research based on reported crime and service call data shows compelling evidence of the impact of the lockdown on VAW. Using police calls for service logs for 15 large cities in the United States, Leslie & Wilson (2020) find that the lockdown due to COVID - 19 has led to a 10.2% increase in domestic violence calls for service or 3.3 additional incidents per day on average. Mohler et al. (2020) also show a significant increase in reported domes tic violence incidents in two large American cities. In Argentina, a recent paper finds a significant increase of 28% in calls to the domestic violence helpline following the mobility restrictions (Perez - Vincent et al., 2020). Using a new victimization survey conducted after the lockdown, the authors also show a positive association between mobility restrictions and IPV, as evidenced by a comparison of women whose partners were exempt from complying with the restrictive orders and women who se partners were not exempt. These grim statistics reflect what has been long - recognized in the literature: there exists a link between crises and gender - based violence (John et al., 2020). For instance, as quarantine measures were imposed during the 2014 - 2016 Ebola crisis, women and girls in West Africa experienced more sexual violence and exploitation (Onyango et al., 2019). Stress, economic anxiety, alcohol, and isolation are some of the factors that explain gender - based violence and are often catal yzed by national crises. The global socioeconomic shock resulting from the pandemic is expected to affect women disproportionately and cut down the progress made over the past decades on gender parity. In coordination with the private sector, some countries are boosting efforts to adapt services to ensure that women can still access them. For instance, in Argentina, Avon has launched a nationwide campaign to raise awareness about domestic violence and promote services oriented towards women. Mobility trends for places of work Daily percentage change for the period Jan 3 - Feb 6 Calls to police's hotline to report GBV: Percent change during the lockdown Source: Google mobility data Source: based on countries’ news and report 2 2 Mexico: K. García, and V. Rojas, “ La Violencia incrementó en la cuarentena más llamadas de auxilio y más búsquedas en Google, ” El Economista . El Salvador: M. Nóchez and V. Gúzman, “ La violencia contra la mujer no encontró amparo durante la cuarentena, ” El Faro. Guatemala: UN. 2020. Guatemala: COVID - 19 Informe de situación No. 04. United Nations. Costa Rica: UNDP. 2020. Análisis trimestral de seguridad ciudadana y respuesta ante el COVID - 19. United Nations Development Programme 5 Our pape r provides a major contribution to the literature in that we use a novel technique for impact assessment, Covariate Balancing Propensity Score (CBPS), to minimize selection bias and address the usual issue of confoundedness. Past studies on the impact of violence have used either an instrumental variable approach (Fajardo - Gonzalez, 2020) or propensity score matching (Morrison & Orlando, 2004). As we discuss in the methodology section, this new propensity score reweighting method, as proposed by Imai & Ratkovic (2014), is an improvement over the usual models used in non - experimental impact assessments of violence given its robustness to misspecification. The paper is organized as follow s. First, we present a systematic review of the evidence on the prevalence, trends, and correlates of VAW in the region. Second, we discuss a theoretical framework linking violence to welfare outcomes and review previous studies on the impact of VAW on wom en and their children's well - being. The section that follows discusses the empirical data and the methodology for a non - experimental impact assessment. Next, we present the results and conclude. VAW in the CID Region As in many other parts of the world, VAW is a prevalent and common issue in the CID region. This section presents an overview of VAW in CID by summarizing what is known about the prevalence and trends of VAW in the region based on the most recent nationally representative population - based data for each country. One of the major issues in a multi - country comparative assessment of the prevalence of VAW is the comparability of data across both settings and studies. Studies such as the WHO Multi - countr y Study on Women’s Health and Domestic Violence and GENACIS (Gender, Alcohol, and Culture) collect comparable data using a standardized questionnaire but have certain limitations. The WHO studies include a limited set of countries in the LAC region, and th e GENACIS is not representative at the national level. We conducted our systematic overview by compiling existing evidence from both surveys and reports from the region. Table A1 provides the sources of the data used. This section focuses on two series of indicators commonly used in measuring violence against women: prevalence of VAW and prevalence of IPV. We define prevalence of VAW as the proportion of women who have reported having experienced at least one act of violence (physical and/or sexual) committ ed by an intimate partner or someone else at any point from age 15. The prevalence of IPV has a similar definition but is restricted to violence suffered at the hand of an intimate partner (current or most recent if not currently in a union). The surveys u sed in this comparative analysis measure physical and sexual violence similarly, making our analysis comparable across countries. Respondents were asked whether they had experienced specific acts of violence, although the surveys did not all measure the sa me acts of violence. 3 Figure 1 shows the most recent VAW prevalence rates, over a lifetime and in the past 12 months, for countries in the CID region. We choose this indicator due to the large overlap between physical and sexual violence in most countries and to facilitate cross - country comparisons since it 3 See Croft et al.’s (2018) Guide to DHS Statistics 7 for an example of the list of acts classified as physical and sexual violence. 6 is the most frequently used standardized measure. The proportion of women in the region who have experienced violence based on gender at least once in their lifetime varies between 7% and 53%. Based on the most recent data, Panama has the lowest rate while Mexico and El Salvador have the highest. For the prevalence of VAW in the past 12 months, we find rates ranging from 5% to 14%. While Panama has the lowest incidence, Mexico, El Salvador, and the Dominican Republic have the highest. Despite consistent efforts to address violence against women in both public and private spheres, the data from nationally representative surveys fail to show a consistent reduction in VAW prevalence over time across countries. Figure A1 shows changes in li fetime prevalence of VAW from previous comparable surveys, indicating longer - term changes in VAW across countries. We restrict this analysis to countries with at least two rounds of comparable datasets collected over the last 20 years. The prevalence of VA W in Honduras increased from 16% in 2005 to 27% in 2007. In Mexico, it increased from 17% to 34% from 2011 to 2016. We also observe an upward trend in Guatemala and the Dominican Republic. The data for Nicaragua and Panama, however, show a downward trend. Intimate partner violence (IPV) captures violence experienced by women in private spheres, at the hands of intimate partners. It is the most common form of gender - based violence and the most pervasive type of violence in the LAC region (Heinemann & Vener, 2006). Figure A2 shows IPV prevalence rates (over a lifetime and in the past 12 months) ranging from 16% to 24% for countries in the region, a level that is below the world average prevalence rate (30%). Women in Panama and El Salvador have the lowest IPV prevalence rate, while women in Haiti and Nicaragua have the highest. Focusing on IPV prevalence over the past 12 months, the rates vary between 5% and 16%. The lowest rate is registered in El Salvador and the highest in the Dominican Republic. From a comparison of VAW and IPV prevalence across countries, we conclude that IPV is indeed a pervasive problem in the region. In Mexico and El Salvador, the issue of VAW is particularly prevalent outside of the home, with high rates of sexual violence against women in working and communal environments. 4 In Mexico, for instance, public transportation is a space where sexual harassment occurs frequently. Accordi ng to the 2016 National Survey of Victimization and Perceptions of Safety (ENVIPE), more than 87% of women feel unsafe when using public transportation in Mexico for their daily commutes. A comparison between lifetime and past 12 months prevalence rates ac ross countries allows us to observe newer term changes, including different life experiences of young women compared to older cohorts. Judging by the minimal difference between prevalence rate for the past 12 months and prevalence over the lifetime, we con clude that VAW, and IPV especially, remains a persistent and chronic problem in the Dominican Republic in particular. This result corroborates the finding that IPV prevalence rates in the Dominican Republic have been on an upward trend in recent years. 4 The Salvador and Mexico questionnaires follow a unique format and collect information on the violence women face in different life environments (family, school, work, and community). The final reports of these surveys show high rates of sexual VAW in co mmunal environments. 7 Femicide is a major issue in the Latin American region, particularly in the Central American countries of Honduras and El Salvador, where it occurs at a significantly higher rate compared to other countries in the region. Figure A4 shows the femicide rates for several countries in the LAC region. El Salvador, Honduras, and Guatemala have the highest rates of these types of violence, while countries such as Costa Rica and Panama have the lowest rates. To reduce these crimes, several countries in the region a re taking important steps, including the classification of femicide as a separate criminal offense and the adoption of minimum and maximum mandatory sentencing. Drawing from the most recent Demographic and Health Surveys (DHS), we conduct bivariate analyse s to examine the relationship between some key socio - demographic variables and VAW prevalence over a lifetime (Figure A5). The choice of the variables considered in this section was informed by a review of the literature on the correlates of VAW. This analysis is based on data from four countries (Haiti, Dominican Republic, Honduras, and Guatemala) for which comparable data exist. The disaggregated VAW prevalence rates wer e averaged across countries. The raw data reveal that women in the region who live in rural areas have a lower average prevalence rate than women living in urban regions. Women below the age of 30 also tend to experience violence at a lower proportion than women above 30. We also observe differences in VAW prevalence rates based on marital status. Women who are separated, divorced, or widowed have a higher average prevalence rate compared to other groups. As opposed to being married, those in a consensual u nion are more prone to experience violence on average. Educational attainment seems to be correlated with the risk of experiencing violence. More years of schooling is associated with lower odds of experiencing violence on average. Wealth status appears to be correlated with the prevalence of VAW as well. Women who live in wealthy households tend to have a lower average incidence rate; although comparing wealth quintiles 1 and 4, there appears to be no difference. Last, the data show that women who are empl oyed have a higher prevalence rate on average than those who are not employed. Figure 1 – Prevalence of VAW in CID Countries (Physical or Sexual Violence) 60 50 40 30 20 10 0 DOM SLV HTI HND GTM MEX NIC* PAN Lifetime Past year Sources: See Table A1 * Only lifetime violence data is available for Nicaragua. Only Physical violence for Panama 52.4 52.7 33.6 27.8 30.3 22.1 4.0 4.2 2.9 9.8 4.0 12.6 7.0 6.7 4.6 1 1 1 1 8 Laws, Policies, and Programs to End VAW In 1993, through the signing of the Declaration on the Elimination of Violence against Women, the United Nations officially recognized VAW as a public concern and asserted women's right to "live a life free of violence." A year later, in Brazil, several co untries across the LAC region adopted the “Inter - American Convention on the Prevention, Punishment, and Eradication of Violence against Women” (also known as the convention of Belém do Pará). In 2015, the United Nations' member states renewed their commitment to end VAW through the Sustainable Development Goals for 2030. 5 Several countries have also agreed to improve their data collection system and measure their advancement using a single indicator (Bott e t al., 2019). More than twenty years after Belém do Pará, the LAC region has made bold advances in the fight to end violence against women and girls, although much remains to be achieved. As of 2020, the 1994 convention has been ratified by 32 out of 33 co untries in the LAC region, except Cuba. Also, almost all countries have enacted national plans to prevent and end violence; some of these plans are designed to address VAW specifically. The international community’s recognition of VAW as a public health is sue ignited a series of legislation in the mid - 1990s aimed at protecting and promoting the rights of women in the region. These laws, often referred to as first - generation laws, guaranteed protection for women against violence experienced in private sphere s (domestic violence). Belize, Costa - Rica, Dominican Republic, Haiti, and Honduras are countries in the region that rely exclusively on first - generation laws (Essayag, 2017). A few years later, a second wave of legislation arose to address the narrowness of first - generation laws. These legislations, known as second - generation laws, extended protection to women against various forms of vio lence experienced in various environments, even forms of violence that were not emphasized in Belém do Pará. 6 Some of these laws consider the diversity of women as a social group, recognizing that violence does not affect all women equally. Today, only Mex ico, El Salvador, Guatemala, Nicaragua, and Panama have adopted these more comprehensive laws on VAW. Eight countries in the region have enacted laws on femicide. These laws come in the form of amendments to current penal codes to legally classify the crime of “aggravated homicide based on gender” (CEPAL, 2015). The Belém do Pará convention also encouraged states to adopt and implement policies to prevent, punish, and eradicate violence based on gender. All countries in the CID region have enacted a national plan and adopted public policies to address the issue of VAW. The national plans often propose activities in the medium and long term to address violence against women and are often constructed as public policy tools with measurable resu lts. States in concert with civil society and the private sector also implement actions to end VAW through diverse programs and services. Most of these programs fall into the categories of prevention, care, or punishment. When it comes to prevention, the m ain programs focus on sensitization campaigns, trainings, and dissemination of values and practices through media. In 5 Pan American Health Organization (PAHO) and WHO member nations also pledged their commitment to end VAW in their respective countries. 6 One such form of VAW is patrimonial violence --- the violation of women’s property rights — which is a continuing issue with significant effects on the lives of women in the region. 9 Mexico, for instance, the Amor, Pero del Bueno program sought to prevent violence in intimate partnerships among adolescents. The program educated middle school and high school students on gender roles and stereotypes and promoted awareness in the educatio nal community. It resulted in a 58% reduction in the percentage of psychological violence perpetrated by young men in relationships and contributed to a reduction in the acceptance of violence and sexism among young people (Sosa - Rubi et al., 2017). On the issue of care, the main strategy is to create service centers for victims of violence. In some countries, integrated services for women are provided in the form of a one - stop service center equipped with medical and psychological care as well as legal aid. This is the case of Ciudad Mujer , a “one - stop shop” model of integrated service delivery in El Salvador that offers multiple services to women, including g ender - based violence support. An impact evaluation of the Ciudad Mujer program shows that it was effective in increasing women’s demand for specialized public services and improving participants’ overall life satisfaction (Bustelo et al., 2019). Another fo rm of care program is establishing 24 - hour national telephone helplines dedicated to victims of violence; Linea Vida in the Dominican Republic is a good example of this type of initiative. As to the theme of punishment, programs mostly focus on the toughen ing of penalties for offenders and reinforcing the justice system to address violence against women more effectively. In Guatemala, for instance, the government has created specialized units to combat the femicide phenomenon. Impact of VAW: Theoretical Framework The impact of VAW on economic variables is a subject that piques the interest of both academics and policymakers. However, the link between gender - based violence and economic development is not always clear - cut, which m ight explain why the issue of violence is often invisible in national strategic plans to boost economic growth. Yet, gender - based violence has implications for economic development through various mechanisms. In a report commissioned by the World Bank, Duv vury et al. (2013) elaborate a model establishing the links between gender - based violence and economic development by discussing some important variables that mediate the impact of violence at the individual level to the macroeconomic level. In the author s’ view, economic outcomes can be affected if there is an overall change in human capital, trauma, or shift in intra - household gender dynamics due to violence. These same pathways can help us understand how violence can impact women’s socioeconomic outcomes and their children’s development. An individual’s human capital, defined as the knowledge and abilities used in the production of goods and services, is often shaped by his or her health con dition and education, both of which can be significantly impacted by previous experiences with violence. Acts of violence can impact an individual’s health (mental and physical) and education, which in turn might impact his or her employment and acquisitio n of skills, among other capabilities. The impact of gender - based violence on the health of victims is well documented in the literature. Studies have shown that victims of violence, IPV in particular, are more likely to use health services (Bonomi et al., 2006) and experience psychological trauma (Swanberg et al., 2005). Violence can also negatively impact the health of the children of victims (Agüero, 2013; Morrison and Orlando, 2004; Bogat et al., 2006). Studies have also shown an impact on the education al performance and 10 behavior of children of victims. Arias (2004) finds that children of women who are victims of IPV are more likely to skip school relative to the children of non - victims. If gender - based violence does not affect an individual's human endowment, it can inflict trauma, which might impact their productivity and/or employment stability. The experience o f violence, either constant or occasional, can instill fear, stress, and anxiety in a person. These psychological outcomes can have serious impacts on the person’s productivity and human capital development. Several studies have shown a correlation between psychological violence and labor market outcomes. Kimerling et al. (2009) show that psychological violence was a stronger correlate of unemployment than physical violence in the United States. Sabia et al. (2013) find that sexual violence is associated with a 6.6% decline in labor force participation in the United States. The last pathway is through changes in intra - household gender dynamics. The exertion of violence against women can change the balance of power within households and consequently put women at a disadvantage (Duvvury et al., 2013). For instance, a loss of bargaining power and decision - making capabilities for women might result in loss of opportunity in the job market and significant household resource all ocation changes. Such shift in power balance could lead to lower investment in children’s education and lower nutrition, considering existing evidence showing that a greater share of household resources under the control of women is associated with greater investment in children's well - being (Rao, 1998). Impact of VAW on Women’s and Children’s Outcomes: Review of the Literature There is a well - known literature on the impact of violence on socioeconomic outcomes. So far, two streams of research have emerged from these studies: one that focuses on estimating the direct economic costs of violence by attaching a monetary value to its impacts, and another that estimates the indirect costs of violence to society. The la tter type of research has focused on analyzing the impact of violence on women’s reproductive, mental, and physical health and socioeconomic outcomes such as educational attainment, labor force participation, and earnings. The consensus in the literature i s that violence has a negative effect on productive outcomes, with consequences being more serious in the area of reproductive health. The impact of gender - based violence on the health of victims is largely documented in the literature. Research has shown that women who are victims of violence have a higher likelihood of experiencing stress, fear, depression, and other psychological trauma (Swanberg et al., 2005). The negative impact of violence on women’s reproductive health is also evident throughout the literature. Women who are victims of violence have a higher risk of adverse pregnancy outcomes (Heise et al., 2002), substance abuse (Heise et al., 1999), cardiovascular disease (Campbell, 2002), among other things. In a study on the impact of VAW in the Latin American region, Agüero (2013) finds a negative association between violence and a series of women’s health outcomes. Women who are victims of violence have lower hemoglobin levels and are therefore more likely to be anemic, an effect that the au thor finds to be more pronounced at the bottom of the distribution. Morrison and Orlando (2004) also find inferior health outcomes for women victims of violence in 11 Peru. Abused women are more likely to have an unwanted last - child, more likely to have a sexually transmitted disease, and more likely to have terminated a pregnancy before term. Regarding the relationship between violence and labor market outcomes, the empirical evidence is mixed. While it is widely argued that violence has a negative impact on labor force participation (Lloyd, 1997; Lloyd & Taluc, 1999; Meisel et al., 2003), some studies have shown a positive correlation between the two variables (Agüero, 2013; Fajardo - Gonzalez, 2020). In the Latin A merican context specifically, Rios - Avila & Canavire - Bacarreza (2017) analyze the heterogeneous effect of intimate partner violence on women’s job exit in Bolivia. They find that violence impacts job exit positively, more so among non - indigenous women. Simi lar results are found in Peru, where being a victim of violence increases the probability of job exit for women (Sierra, 2015). On the other hand, studies such as Morrison & Orlando (1999) and Agüero (2013) find intimate partner violence to be positively a ssociated with labor force participation; in other words, abused partnered women are more likely to work than non - abused partnered women. Similarly, using the most recent DHS for Colombia, Fajardo - Gonzalez (2020) finds a positive relationship between domes tic violence and women’s employment even after correcting for endogeneity. Agüero (2013) hypothesizes that part of the effect of violence on women’s labor supply might be due to changing marital status (from marriage to divorce). Using mediation analysis, Fajardo - Gonzalez (2020) shows that the positive association between IPV and women’s employment could be explained by a desire to enhance their bargaining power in the hope of exiting abusive relationships. When it comes to the impact of VAW on workers' pro ductivity, the literature has shown a negative correlation (Swanberg et al., 2005; Reeves, 2004). Existing research has shown that victims of violence often experience a higher level of distraction and are more frequently absent from work due to physical a buse or threat of violence (Logan et al., 2007). The evidence on the impact of violence on employment instability is also clear. Victims who experience gender - based violence have higher rates of absenteeism and tardiness with significant impact on job perf ormance. Women who are victims of gender - based violence are more likely to lose their jobs and experience a higher turnover than non - victims (Bell, 2003; Meisel et al., 2003; Swanberg et al., 2005). VAW has also been shown to have a negative impact on wome n’s earnings through missed paid work (Duvvury et al., 2012). Indeed, VAW comes at a high cost to businesses and the economy overall. A series of recent studies have empirically assessed the business costs of VAW in several South American countries. In Per u, it is estimated that IPV is responsible for 70 million lost workdays per year, which amounts to approximately 6.7 billion dollars or 3.7% of the country’s GDP (Vara Horna, 2013). In Paraguay, IPV costs businesses an estimated 734.9 million dollars per year, equivalent to 6.46% of GDP (Brendel & Heikel, 2015). VAW can also have intergenerational effects; when women experience violence, their children also suffer. Several studies have examined the impact of exposure to violence on the health of children. Most of them have concluded that children exposed to violence are less healthy than those who are not exposed to violence. For instance, Agüero (2013) finds that children whose mothers are victims of violence fare worse in health outcom es before and after birth. While in utero, children of abused women are less likely to have the required four or more prenatal visits. 12 Once born, they are less likely to be vaccinated, more likely to have had diarrhea in the past 15 days, and more likely to be underweight compared to children of non - abused women. In Peru, Morrison and Orlando (2004) also find that the children of abused w omen have a higher likelihood of suffering from diarrhea and fare worse in anthropometric measures. They were, however, more likely to be vaccinated compared to the children of non - abused women. In the United States, Bogat et al. (2006) find children of victims to be more prone to experiencing trauma from hearing and witnessing abuse. There is also evidence that exposure to violence can affect the educational attainment and behavior of children. Arias (2004) finds that the children of women who are victims of violence are more likely to skip school and suffer poorer health than the children of non - victims. Research in Nicaragua has also shown that the children of victims are more likely to repeat a school year and drop out of school early (Morrison and Orlando, 1999). Absenteeism from school is also an issue among children of victims. Emery (2011) studies children in one American city and shows truancy to be higher among children of victims. It has also been shown that c hildren who are exposed to violent behaviors have a greater tendency to imitate and reproduce the cycle of violence witnessed (Enlow et al., 2012). Methodology: Non - Experimental Impact Assessment Empirical studies aiming to assess the indirect costs of gender - based violence often resort to a comparison between a group of women who have suffered from violence to a control group (women who have not suffered from violence) in a non - experimental format , since a randomized controlled experiment in this context would be both impractical and unethical. In such a case, the statistical difference in a particular outcome between the control and treatment groups would inform whether there is an impact. These s tudies, however, often suffer from the issue of selection bias ever - present in non - experimental impact assessments and evaluation studies. The sample of women who report having experienced violence is often dissimilar to the control group in terms of chara cteristics, thus rendering the effect of violence hard to isolate. Another concern in these types of exercises is the presence of endogeneity between the outcome variable and VAW. For instance, considering the outcome variable “participation in the labor m arket," violence perpetrated against a woman within the household might force her to enter or leave the labor market. By the same token, women’s participation in the labor market might itself be a cause of violence (Morrison et al., 2007; Hjort & Villander , 2012). One way of addressing the reverse causality and selection bias issues is to use an instrumental variable (IV) approach, often through a two - stage linear probability model (see Fajardo - Gonzalez, 2020). The instrumental variable must satisfy the exc lusion principle, meaning that it must not directly influence the outcome under study but must be correlated to the treatment or grouping variable. However, the IV method must be used with extreme caution given the practical problems one can encounter in i dentifying valid instrumental variables (Crown et al., 2011). More rigorous studies make use of the statistical technique of matching to address the sample selection issue (Morrison and Orlando, 2004). The Propensity Score Matching (PSM) 13 technique allows researchers to construct sets of individuals from treatment and control groups that share similar characteristics. Using scores based on the probability of suffering violence, each individual in the treatment group is matched with an individual in the control group with the closest propensity scores. PSM has two important properties. It satisfies the “common support” condition, which is necessary for an appropriate comparison of treatment and control groups. Given its non - parametric nature, it also bypasses the complications of the choice of functional form and the complexities of using instrumental variables (Sánchez & Ribero, 2004). The PSM method is not without criticism. It often requires a larger sample, as observations that fail to be matched are often excluded, and can be sensitive to omitted variable bias. Inverse probability weights based on propensity scores can also be used to address endogeneity and selection bias concerns. This technique, known as propensity score reweighting (PSR), allows a researcher to create balanced treatment and control groups that simulate random allocation of subjects similar to the PSM method. The PSR has a major advantage over the PSM method in the sense that it retains all the observations, which helps maintain statistical power to detect a treatment effect (Stone and Tang, 2013). One of the goals of this chapter is to estimate th e impact of violence on a series of key outcomes related to women's reproductive health and socioeconomic condition. We intend to assess differences in outcomes between two groups of women: those who have experienced IPV in the past 12 months and those who have not. Also, we will assess differences in outcomes related to children’s health and education between two groups of children: those whose mothers have experienced IPV in the past 12 months and those whose mothers have not. 7 To deal with the issues of confoundedness, we use a novel methodology, Covariate Balancing Propensity Score (CBPS), as proposed by Imai & Ratkovic (2014), to produce comparable estimates of the impact of violence in four countries in the region: Haiti, Guatemala, Honduras, and the Dominican Republic. Standard propensity score models maximize the likelihood function's empirical fit to optimize treatment status prediction, but covariate balance is not always addressed. CBPS optimizes the covariate balance while modeling the t reatment assignment, allowing near - perfect covariate balance between control and treatment groups. The CBPS is a significant improvement over the PSM and standard PSR methods, given its robustness to the propensity score model's misspecification. In a simu lation study, Imai & Ratkovic (2014) find that CBPS performs better than other propensity score models in terms of bias and mean square error. Therefore, this nascent method has been used in several studies in various applied disciplines to deal with confoundedness when assessing causal effects using observational data (see Albanese et al., 2021; Ehrenthal et al., 2016; Vandecandelaere et al., 2016). Our first goal is to make the control and treatment groups look similar, thus comparable, over a series of control variables using weights based on propensity scores. After reweighting, under the assumption of limited omitted variable bias, the difference in a particular outcome (i.e., participation in the labor force) between the two groups is the average treatment effect (in our case, the effect of violence on the outcome being assessed). In other words, given that the average 7 For the empirical analysis, given that most of the outcomes are linked to household decision - making, we are using IPV p revalence instead of the broader VAW measure as the grouping variable to better capture the effects of violence. IPV is the most common and pervasive form of VAW in the LAC region (Heinemann & Vener, 2006). 14 treatment effect is of interest, we can weigh the control group observations (women who have not experienced violence) such that their (weighted) covariate distribution matches with that of the treatment group (women who have experienced violence). Using CBPS, the average treatment effect can be obtained through these steps: Step 1: A discrete choice model (probit) is run: Prob (tvar = 1 | X) = invprobit (X * b) where tvar is the treatment variable (having experienced violence or not), X is a matrix of control variables, and b is a vector of coefficients to be estimated. Unlike other propensity score methods, CBPS yields the probit coefficients (b) that produce the best balance on matching variables while modeling the treatment assignment. Step 2: Wei ghts are constructed as follows: Weights for the treatment group: 1/p Weights for the control group: 1/(1 - p) Where p is the predicted value (the propensity score) based on the model in step 1 Step 3: We regress the outcome variable, using the weights, on t he treatment variable for evidence of impact. Y = tvar * b Y is the outcome variable, and b is the average treatment effect (the effect of violence on the particular outcome) Data For the empirical analysis, we draw from the most recent DHS for four countr ies in the CID region: Dominican Republic, Haiti, Guatemala, and Honduras. The DHS gather demographic and socioeconomic information for women and children and are representative at the national level. They collect information at both the household and individual levels. In the 1990s, the Measure DHS program added a specific module on domestic v iolence in a few countries’ surveys to better understand the link between violence and health outcomes. The module is answered by women between the ages of 15 and 49. The DHS use the modified Conflict Tactics Scale (CTS) to measure intimate partner violenc e. The CTS is the most internationally accepted method of measurement of gender violence (Morrison et al., 2007). It is unique because it consists of questions on specific acts of violence ranging from less severe to severe, thus reducing the probability o f self - censorship, an incident that often occurs when women are asked directly about previous experiences with violence. We restrict our analysis to four countries for which recent DHS data are readily available. The sample sizes range from 4,322 in Haiti to 12,494 in Honduras. Our samples consist of women 15 who were selected for the domestic violence module. One of the D