Household sanitation and non-partner sexual violence against women Evidence from India

Master Thesis for the Double Degree Program

M.Sc. International Economics and Business, Rijsunivserseit Groningen M.A. International Economics, Georg-August-Universität Göttingen

Household sanitation and non-partner sexual violence against women

Evidence from India

Submitted by: Submission date:

Johanna Gradl June 18th, 2019

Boterdiep 86A 9712 LS Groningen

j.m.gradl@rug.student.nl Supervisors:

Abstract

Anecdotal evidence suggests that Indian women who have to leave their household for toilet use are exposed to an increased risk of non-partner sexual violence. Utilizing data from the latest two rounds of the Demographic Health Survey, a logistic probability model is applied to calculate whether women who lack access to private sanitation have a higher risk of experiencing rape than women with household toilet facilities. The results suggest that a link from lacking a household sanitation facility to having a higher risk to become a victim of non-partner sexual violence might indeed exist in rural areas. In line with the hypothesis, no effect is found in the urban sample. This implies that sanitation planning should incorporate gender-responsive sanitation solutions.

Keywords: Non-partner Sexual violence, Outside defecation, Demographic Health Survey, India

Acknowledgment

Table of Contents

1. Introduction ... 1

2. Conceptual Framework ... 3

3. Context ... 5

4. Data and Method ... 8

1.

Introduction

Violence against women (VAW) is one of the most widespread violations of human rights and a worldwide public health concern. Defined as any act of “gender-based violence that results in or is likely to result in physical, sexual or psychological harm or suffering to women […]”, VAW is a problem in all societies and cuts across lines of income and culture (United Nations 2015, p.1). While most of VAW is intimate-partner violence, non-partner sexual violence (NPSV) must not be neglected. Seven percent of women worldwide are estimated to have experienced NPSV at least once in their lifetime. NPSV can cause serious physical, mental, sexual and reproductive health problems, such as injuries, unintended pregnancies and abortions, suicide, depression or post-traumatic stress, with social and economic costs for the whole society (United Nations 2015). While gender inequality and society norms that ascribe a lower social status to women and accept VAW are considered the root cause of VAW, a small number of studies seeks to identify the influence of neighborhood and community factors on VAW (Frye and O’Campo 2011, Parks 2014, Kiss 2015).

In this paper, I investigate one potential community risk factor for NPSV: lacking access to private household sanitation which implies using shared toilet facilities or practicing open defecation. Worldwide, 892 million people, 12% of the world population, still practice open defecation and India is home to more than half of them (WHO/UNICEF 2017).1

The high prevalence of open defecation combined with low gender equality and anecdotal evidence on NPSV when traveling to open defecation sites makes India an important context to study the link between access to sanitation and NPSV. The power imbalance between gender in India also reflects in the sanitation reality. Open defecation is stigmatized for women and cultural norms make them seek the privacy of the dark to go outside and relieve themselves. This stigma is less prevalent and restricting for men. Subsequently, many women urinate and defecate only once or twice a day, in dusk or dawn and have to travel long distances to find privacy (Sahoo et al. 2015). Those women are prisoners of daylight and this has serious health consequences.2 _{Beside physical health impacts,}

women have additional concerns and face psychosocial stress that comes along with lacking a safe sanitation facility. In particular fear of sexual violence and harassment as well as experienced incidents of NPSV when being outside for toilet use adds to the special burden for women. Nevertheless, only

1 _{Unimproved sanitation facilities in general and particularly the practice of open defecation is contributing significantly} to the global burden of disease. Among the diseases transmitted through poor sanitation are trachoma, worm infections and diarrhea. The latter causes 11% of child deaths worldwide. If not fatal, the diseases also entail severe consequences for childhood growth and cognitive development (WHO/UNICEF 2013).

little empirical literature quantifies the effect of poor sanitation facilities on the risk of NPSV. Tragically, media reports since 2013 make the link between open defecation and NPSV more relevant than ever.3

In this vein, I seek to add to the existing literature on sanitation related NPSV. Understanding the drivers and risk factor of NPSV and identifying sanitation situations in which women are particularly vulnerable for NPSV is critical to generate sustainable sanitation solutions and to reduce VAW. By utilizing a large cross-sectional dataset, the Demographic Health Survey (DHS), I calculate the odds ratios of experiencing NPSV for women who have to defecate outside the household in comparison to women with access to a household toilet. I find that outside defecation is increasing the risk of experiencing NPSV in rural areas, but not in urban. At the same time, open defecation has no measurable impact on intimate partner sexual violence (IPSV), which renders the hypothesized causal link more plausible and speaks against unobserved confounding characteristics of women influencing both, their likelihood of having no sanitation facility and of being a victim of violence.

The remainder of this paper is structured as follows: In Section 2, I conceptualize my hypothesis by analyzing the existing literature on the interrelation between household sanitation access and NPSV. Section 3 provides relevant country information for the research context. I describe the data I used in Section 4, followed by the construction of variables, the empirical model and descriptive statistics. In Section 5, I report and discuss the estimation results and provide alternative estimations with different variable specifications. Section 6 concludes and provides policy implications.

2.

Conceptual Framework

A number of household surveys throughout India report that women face additional psychosocial stress when using an open defecation site (Khanna and Das, 2016 in rural Uttar Pradesh, Sahoo et al. 2015 in rural Odisha, Hirve et al. 2015 in rural Maharashtra, Nallari 2015 in urban Karnataka, and Truelove 2011 in urban slums in Delhi). In all those surveys, the majority of respondents has a heighted fear of rape and sexual assault associated with open defecation.

Only a limited number of studies use secondary data to link reported NPSV with the lack of sanitation in developing countries. For an urban township in South Africa, Gonsalves et al. 2015 linked the time spent outside to find a toilet and the number of facilities in the area with sexual assault applying a mathematical model. They find that the number of sexual assaults could be lowered by increasing the number of toilet facilities. A quantitative study with the Kenyan DHS data from 2008/09 established a significant relationship between open defecation and non-partner violence. The effect was reported particularly strong in disorganized communities (Winter and Barchi 2016).

Jadhav et al. 2016 used the Indian DHS dataset from 2005/06 to investigate the relationship between the access to household sanitation and NPSV in India. They find a significant association between open defecation and NPSV. According to Jadhav et al., women who practice open defecation have 2.14 times the risk of NPSV of women who use a toilet. The study of Jadhav et al. will serve as a first reference point for this paper and I will further build upon it. Jadhav et al.’s research is an important first attempt to explore the predictors of NPSV in India with DHS data. The study has some important drawbacks that I seek to address.

First, unlike Jadhav et al. I will use a dataset from 2015/16. To my knowledge, no paper so far investigated the link between NPSV and household sanitation with this data set. While the effect of a lack of sanitation on NPSV is not excepted to be per se period specific, it might be that reported cases in the media and the nationwide outrage from 2012 onwards4 _{encouraged women to disclose their own}

experiences with NPSV when interviewed. Further, sanitation campaigns are still in progress and the Government of India (GoI) planned to reach a milestone of open defecation free India in 2019. Given those efforts, the sanitation situation could have been subject to changes since 2005/06. The relevance of the topic requires re-testing the relationship with more up to date data. At the same time, this allows to compare results from data of 2005/06 with those from of 2015/16.

Moreover, Figure 1 shows large differences between the urban and the rural sanitation situation in India. Generalizability of results might be limited when considering India as a homogenous setting and disregard the fundamental differences between areas of residence. This distinction might be particularly relevant in the Indian context (as further described in Section 3) and has not been considered in the literature so far.

The DHS data sets that I use do not allow to control for several potential driver of NPSV. However, an important circumstance is whether a woman is employed or not. Women in the workforce also need to travel to work, are exposed to the environment and can barely be accompanied. Jadhav et al.’s model did not consider current employment. Moreover, sexual violence at the work site is possible and this setting might even facilitate violence due to the dependency of women from employment. Also, the number of women in the household has not been considered so far. This might influence whether a woman travels to a sanitation facility on her own.

Further, I introduce the term outside defecation. Outside defecation, in contrast to open defecation that is used in other studies, describes not only walking outside to an open defecation site, but also walking outside to a shared and public facility. In particular in the Indian setting where toilets are not desired close to houses, traveling to such a facility, especially at night, might be a risk factor. Research so far has not considered whether a toilet facility is shared or private. Applying the term outside defecation for all forms of defecation outside the household seems more appropriate in this context. Given the aforementioned literature and in particular the Indian setting, that will further be explained in Section 3, I hypothesize that women who practice outside defecation have higher odds of experiencing NPSV than women who use a private household facility. I also anticipate that NPSV in urban and rural is to some extent driven by different risk factors. In particular rural India remains a globally unique sanitation conundrum. Thus, I expect the lack of a household facility to be a prevalent risk factor for NPSV in rural but not necessarily in urban areas.

3.

Context

The combination of increased exposure to feces due to high population density and the large number of people still defecating in the open makes India a sanitation crisis hotspot. Seven % of urban and 56% of rural dwellers still practice open defecation. The Indian rate of 40% (524 million people) of the total population practicing open defecation by far exceeds those of most other developing countries (WHO/UNICEF 2017). Figure 1 shows the development of India’s open defecation rates in comparison to the neighbors Bangladesh, China and Pakistan. While open defecation is often seen as a function of poverty, high illiteracy rates and scarce water resources, India offers rapid economic growth, increasing literacy rates and improved water availability. Nevertheless, each year India’s fraction of the globally remaining people practicing open defecation is increasing (Coffey et al. 2017a). This is foremost driven by the 67% of Indians living in rural areas. The number of rural people with latrine only increased modestly from 22 % in 2001 to 31 % in 2011 - despite the GOI’s latrine construction efforts since 1986 (Census India 2011).5

Figure 1: Open defecation rates as % of population 2000-2015

Source: WHO/UNICEF Joint Monitoring Programme (JMP) for Water Supply and Sanitation.

Note: The average percentage of people defecating openly for all lower middle-income countries is 24 %. Countries with higher open defecation rates than India are Sub-Saharan African countries such as Burkina Faso, Benin, Chad, Eritrea, Madagascar or Niger.

5 _{In 1986 the GoI launched the first large-scale sanitation initiative and households were subsidized with hardware to} generate demand for toilet construction. After this initiative failed to address the behavior change from open defecation to latrine use, it was restructured in 1999 as the Total Sanitation Campaign (TSC). In 2012, the TSC was renamed Nirmal Bharat Abhiyan (NBA, Clean India Campaign), aiming to achieve 100% sanitation access for rural households by 2022. The community-led-approach was intensified, and cash incentives were increased, as well as extended to above poverty households, which were still identified poor (Routray et al. 2017). In 2013, Prime Minister Narendra Modi announced, “Toilets first, temples later” and relaunched the TSC as the Swacch Bharat Mission (SBM, Clean India Mission) in 2014 to achieve an open defecation free India by 2019.

Rural India Urban India India 0 10 20 30 40 50 60 70 80 90 100 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 P E RC E N T O F P O P U L A T IO N

The GoI fails to acknowledge that high sanitation coverage does not mean that people stop open defecation (Barnard et al 2014, Hueso and Bell 2013). While constructing toilets is essential, it does not automatically bring along the behavioral change from open defecation to actually using and maintaining the toilets. In particular in rural India, defecating openly has a long history and is established as a socially accepted habit in areas where everybody practices it. It can be seen as an institution and abandoning it requires a change in mind-set as this quote of a man in Odisha demonstrates: “If [open defecation] was good enough for the Maharajas, it’s good enough for me” (Dickinson et al. 2009 p. 2). Beside not considered embarrassing - at least not among men - it is often thought to be healthier using “open sky latrines” (Dickinson and Pattanayak 2009, p. 2). Also, missing trust in the health-benefits of latrine-use – that are not directly visible - was recorded (Coffey et al. 2017a). Furthermore, norms of purity and pollution determine behavior in India such as supposedly in no other country. Coffey at al. 2017 refer to research carried out by Khare (1962), who describes the distinction between ritually polluting and physically dirty. In many households, latrines and feces are considered ritually polluting and defecation in general, openly or in latrines, should be carried out far from the house to keep it ritually and physically pure. People also reject the GoI’s pit latrines due to concerns about emptying the pit one day. This is rooted in Indian-specific beliefs, values and norms about purity, pollution, caste and untouchability.6 _{Nevertheless, there does not seem to be an absolute}

disdain towards latrines. Expensive latrines with large, cemented underground tanks are mostly excluded of being seen polluting. Western flush toilets, which are connected to canalization are considered features of wealth and good social standing. However, a lot of rural Indians rather defecate outside than using the affordable latrines with small soak pits that are promoted through the GoI during the sanitation campaigns (Coffey et al. 2017a).

Beside the lasting high prevalence of open defecation, India has gained international media attention due to several cases of sexual violence against women in the recent years.7 _{Several local and}

international media reports linked lacking sanitation with sexual VAW in India. A global media

6 _{The lowest caste-group, known as Dalits or Untouchables, is socially marginalized and oppressed as they are considered} permanently polluted. Dalits are traditionally obliged to do dirty work for upper-caste households, which also includes manually cleaning feces. Under the pretext of pollution, Dalits are excluded from other employment, schools and temples. However, there is slow progress to the better for Dalits. They started refusing to empty others’ pits in their pursuit for equality and dignity. Further, employing ‘manual scavengers’ was prohibited in all states in 2013 (Coffey et al. 2017a, Human Rights Watch 2014). Cleaning twin-pit latrines, which means that one pit can compost until emptying is harmless while the other pit is in use, would be excluded from the prohibition. Contrary to the GoI’s statement of promoting pit latrines for decades, the SQWAT survey showed that this technology is rather rare (Coffey et al. 2017b).

outrage started in 2013 with the case of two teenage cousins in rural Uttar Pradesh who were raped and lynched when they were on their way to an open defecation site (The Guardian 2014). Adding to the problem is how Indian police and government officials dealt with rape cases in the past. Many were accompanied by outrageous comments of politicians.8 _{VAW, gender inequality and women’s}

low social standing thus have even been legitimized from public authorities.Consequently, the low social standing of women and the patriarchic structures in the country make inadequate sanitation even more problematic.

8_{“These kind of incidents [rape] happen accidentally.“ (Home Minister of Chhattisgarh); “Boys and girls love each other,}

4.

Data and Method

This paper utilized cross-sectional data from the 2015-2016 India DHS, also known as the fourth version of the National Family Health Survey (NFHS 4). The International Institute for Population Sciences collected information from a nationally representative sample of 616,346 households from January 2015 until December 2016 with a series of standardizes questionnaires. The respondents for the survey were identified by a two-stage cluster sample.9 _{Household surveys report}

physical, social and economic household characteristics. Individual interviews that specifically focus on health, livelihood and experiences of certain social issues have been completed with a subsample of men as well as with 699,683 women age 15-49. Within this paper I solely use the women data set which has been merged to the respective household characteristics from the household survey.

One woman per household was randomly selected for a domestic violence (DV) module and asked to answer 40 additional questions about different types of violence subsequent to the standard questionnaire. All interviewers were obliged to solely ask those questions in privacy. Only one woman per household was included to further assure that no other household member would know about the content of the questions asked. At this point it should be recognized that the DV module is aimed at monitoring domestic violence. Non-partner and non-domestic violence were not the priority but have also been interrogated to some extent. Further drawbacks of the survey design will be discussed later on.

Additionally, I use the Indian DHS round from 2005-2006 (NFHS 3) as it includes a similar DV module. Certain differences in the design of the data sets do not allow the exact same specifications within the estimated regressions with 2015/2016 data.10 _{Nevertheless, it allows for comparison over}

time. The composition of the final samples and how states have been assigned to regions can be found in Appendix 1 and 2.

9 _{The 2011 population census served as the sampling frame. The sample was stratified into urban and rural, at the first} stage the primary sampling units (clusters) were selected, in the second stage 22 households per cluster were chosen with equal probability. Only pre-selected households were interviewed. A subsample of eleven households were selected for the women’s questionnaire and all women 15-49 who are household members or spent the night before the interview in the household were eligible to be interviewed.

10 _{NFHS 3 does not contain districts, only city size, which NFHS 4 does not contain; NFHS 3 only includes 29 states and} union territories, while NFHS 4 contains all states and union territories.

4.2 Empirical Model

A logistic probability model is implemented to estimate the risk for women to experience non-partner sexual violence when lacking household sanitation facilities in comparison to women whose households do have sanitation facilities.11 _{The main empirical model is as follows:}

𝑁𝑜𝑛 − 𝑝𝑎𝑟𝑡𝑛𝑒𝑟 𝑣𝑖𝑜𝑙𝑒𝑛𝑐𝑒_/ = 𝛼_/+ 𝛽₄(𝐿𝑎𝑐𝑘 𝑜𝑓 𝑡𝑜𝑖𝑙𝑒𝑡 𝑓𝑎𝑐𝑖𝑙𝑖𝑡𝑦)_/+ 𝛾₄(𝑊𝑜𝑚𝑎𝑛 𝑐ℎ𝑎𝑟𝑎𝑐𝑡𝑒𝑟𝑖𝑠𝑡𝑖𝑐𝑠)/+

𝛾_@(𝐻𝐻 𝑐𝑜𝑛𝑡𝑟𝑜𝑙𝑠)/+

𝛾B(𝐺𝑒𝑜𝑔𝑟𝑎𝑝ℎ𝑖𝑐 𝑐𝑜𝑛𝑡𝑟𝑜𝑙𝑠 )/+ 𝜀/ (1)

where non-partner violence is the incidence of NPSV within the last 12 months. Lack of toilet facility is a binary variable that is generally 1 if the toilet facility is inadequate or missing in the household and 0 otherwise. Several demographic characteristics, household controls and geographic dummies are incorporated. 𝜀/ is the error term.

In order to identify whether the respondent had experienced NPSV, the following questions have been asked within the DV module. All selected women have been asked: “At any time in your life, as a child or as an adult, has anyone ever forced you in any way to have sexual intercourse or perform any other sexual acts when you did not want to?”

All never-married women who replied with yes, where then asked: “In the last 12 months, has anyone physically forced you to have sexual intercourse when you did not want to?”, while ever-married women were asked: “In the last 12 months, has anyone other than (your/any) husband physically forced you to have sexual intercourse when you did not want to?”

A NPSV variable was coded 1 for all ‘yes’ answers to the last two questions and 0 otherwise. Thus, the NPSV variable describes whether a woman has experienced sexual violence within the 12 months prior to the interview by anyone than the partner. Less than 0.6% of respondents (433 women) refused to answer one of the questions. No follow up question about the identity of perpetrator referred to in the last two quoted questions has been asked in the DV module. The sole information about the

identity of any other perpetrator than the husband is collected for the first forced sexual act.12 _Also,

the survey does not include any information whether the (sexual) violence occurred within the house, on the way to the sanitation facility or elsewhere. Simultaneously, it cannot be ruled out that the NPSV is still domestic violence, thus executed by someone living in the same household who is not the husband.13

The two violence measures, NPSV and IPSV are restricted to the last 12 months before the interview for two main reasons: First, the temporal ordering ensures that NPSV and the type of sanitation can be linked despite having cross-sectional data. Second, due to the missing information on the perpetrators, the question about sexual violence within the last 12 months is the only sexual violence related question in which the husband or partner is explicitly excluded.

The information about the toilet facility is obtained through the following two questions from the household questionnaire: “What kind of toilet facility do members of your household usually use?” and “Do you share this toilet facility with other households?”14

The variable outside defecation is coded 1 if either “no facility/bush” or a shared facility has been reported. In this context outside defecation describes the necessity to leave the own household to use a toilet facility.

I control for characteristics of the women, such as age, highest educational level, the relationship status and whether a woman is currently employed or not. Available information on religion and caste is also incorporated. I also consider the household’s economic status. However, the wealth index provided by the DHS is a composite measure of the household’s cumulative living standards and consists of several weighted household assets, among others the toilet facility. By controlling for an unadjusted wealth index, I would control for the sanitation facility twice. Thus, I could either alter the wealth index and exclude the toilet facility, or control for other household assets that are also contained in the wealth index. Within the main regression, I do the latter by including electricity and roofing type as done by Jadhav et al. (2016).Further, I control for regional ‘fixed effects’ by including region dummies. Those dummies absorb unobserved or observed effects particular to each region. Although data was collected in two consecutive years, year-fixed effects are not included, and I do not assume that the respective year is a potential confounder. The data collection was done state wise, half of the states were included in the first, the other in the second round. That means that the year effect would be simply driven by the states that were included in that specific year.

The general difference in the social and economic setting of urban and rural means different realities for women. The presence of large differences in those settings might affect the generalizability of the conclusions that can be drawn from the regression results. As hypothesized, I expect the areas of residence to differ for the research question. Therefore, I first, include the area of residence in the full model to check whether it might be a driver of NPSV. Second, I run separate regressions on the urban and the rural subsample. Summary statistics of all control variables can be found in Appendix 4 and 5.

Women who practice open defecation might share other characteristics that make it more likely for them to live in households that lack sanitation facilities and simultaneously experience sexual violence (Jadhav et al. 2016). Therefore, I test in a second model whether outside defecation affects the risk of IPSV in the prior 12 months. If there was an effect of outside defecation on IPSV, the channel of leaving the house in the dark to walk to public facilities or unlit open defecation sites would be statistically questioned. One could assume that unobserved factors influence both, living in a household without sanitation facility and being exposed to sexual violence. The empirical model is as follows:

𝐼𝑛𝑡𝑖𝑚𝑎𝑡𝑒 − 𝑝𝑎𝑟𝑡𝑛𝑒𝑟 𝑠𝑒𝑥𝑢𝑎𝑙 𝑣𝑖𝑜𝑙𝑒𝑛𝑐𝑒_/ = 𝛼_/+ 𝛽₄(𝐿𝑎𝑐𝑘 𝑜𝑓 𝑡𝑜𝑖𝑙𝑒𝑡 𝑓𝑎𝑐𝑖𝑙𝑖𝑡𝑦)_/+ 𝛾₄(𝑊𝑖𝑓𝑒 𝑐ℎ𝑎𝑟𝑎𝑐𝑡𝑒𝑟𝑖𝑠𝑡𝑖𝑐𝑠)/+

𝛾_@(𝐻𝐻 𝑐𝑜𝑛𝑡𝑟𝑜𝑙𝑠)_/+

𝛾B(𝐺𝑒𝑜𝑔𝑟𝑎𝑝ℎ𝑖𝑐 𝑐𝑜𝑛𝑡𝑟𝑜𝑙𝑠 )/+ 𝜀/ (2)

4.3 Descriptive Statistics

Figure 2 offers a comparison between the data from 2005/06 and 2015/16 by region and residence area. In all regions the percentage of people practicing outside defecation decreased at least to some extent. However, women’s access to household toilet facilities still varies widely across regions. Also, the toilet setting remains very differently across urban and rural. In the latter, the share of people practicing outside defection is generally higher and ranges from 76% (of the rural population) in North-Central to 23% in the North-East. The North-East is the only exception where a larger percentage practices outside defecation in urban than in rural in 2015/16. Table 1 and 2 show the summary statistics of the sanitation pattern and the composition of outside defecation. In NFHS 4, 35.5 % of the total sample practice open defecation while 9.1 % use a shared facility, accordingly 43.6 % go outside their household for defecation. In NFHS 3, 53.9% reported open defecation and 11.6 % a shared facility which results in 65.5% of the full sample to practice outside defecation. Thus, outside defecation in the full sample is primary driven by open defecation. Therefore, region wise prevalence and development of open defecation between NFHS 3 and 4 is demonstrated in appendix 3.

Figure 2: Outside of household defecation comparison NFHS 3 and 4, by region and area

Note: Figure is based on weighted samples. NFHS 3 is DHS round 2005/2006 while NFHS 4 is DHS round 2015/2016. NFHS 3 does not include union territories except of Delhi. Regions are divided as follows: North-east: Arunachal Pradesh, Assam, Manipur, Meghalaya, Mizoram, Nagaland, Sikkim, Tripura; West: Goa, Gujarat, Maharashtra, Dadra and Nagar haveli, Daman and Diu; South: Andhra Pradesh, Karnataka, Kerala, Tamil Nadu, Telangana, Puducherry, Andaman and Nicobar islands, Lakshadweep; North: Haryana, Himachal Pradesh, Jammu and Kashmir, Punjab, Rajasthan, Uttarakhand, Delhi, Chandigarh;

North-central: Chhattisgarh, Madhya Pradesh, Uttar Pradesh; East: Bihar, Jharkhand, Odisha, West Bengal. NFHS 3 does not

contain the union territories except of Delhi.

0 20 40 60 80 100 O ut side de fe ca tio n %

North-East West South North North-Central East

NFHS 3 NFHS 4 NFHS 3 NFHS 4 NFHS 3 NFHS 4 NFHS 3 NFHS 4 NFHS 3 NFHS 4 NFHS 3 NFHS 4

In terms of open defecation, differences between urban and rural are even more significant (See figure in appendix 3). Open defecation is the reason for high outside defection rates in rural areas, whereas outside defecation in urban is dominated by the use of shared facilities. Also, urban areas within India remain heterogenous in 2015/16. The urban North-east can be considered largely open defecation free (in NFHS 3 and 4), while an open defecation rate of 17 % in the East and North-central is still high for an urban setting given the Indian urban average of 9.3 % (NFHS 4). Both figures (figure 1 and appendix 3) suggest that improvements in the toilet setting between both rounds have been only moderate. Especially, the fact that the ten years and several GOI Clean India Missions and campaigns lie in between the two survey rounds, demonstrates that India’s toilet crisis is highly persistent and differently dominant within India.

Table 1: Descriptive statistics: Full, Urban, Rural Sample; NFHS 4 2015/2016

Table 2: Descriptive statistics: Full, Urban, Rural Sample; NFHS 3 2005/2006

Sanitation

Outside defecation 0.655 (0.475) 0.398 (0.489) 0.780 (0.415) Open defecation 0.539 (0.498) 0.161 (0.367) 0.722 (0.448) Using a shared/public facility 0.116 (0.320) 0.237 (0.425) 0.058 (0.234)

Sexual violence

Non-Partner Sexual Violence within

last 12 months 0.002 (0.040) 0.001 (0.035) 0.002 (0.043) Intimate-Partner Sexual Violence

within last 12 months* 0.027 (0.161) 0.018 (0.133) 0.030 (0.171)

Note: Tables 1 and 2 are based on weighted samples; always Min=0 and Max=1; IPSV is measured among married women only (NFHS 4: 58,719 observations; NFHS 3: 60,516 observations).

Variables Full Sample _{N= 75,018} Urban Sample _N=22,277 Rural Sample _N=52,741

Sanitation

Outside defecation 0.436 (0.496) 0.233 (0.423) 0.540 (0.498) Open defecation 0.345 (0.475) 0.093 (0.290) 0.473 (0.499) Using a shared/public facility 0.091 (0.289) 0.139 (0.346) 0.067 (0.250)

Sexual violence

Non-Partner Sexual Violence within

last 12 months 0.004 (0.061) 0.006 (0.076) 0.003 (0.052) Intimate-Partner Sexual Violence

within last 12 months* 0.012 (0.106) 0.012 (0.108) 0.011 (0.105)

Generally, reported NPSV is relatively rare in the Indian setting. In NFHS 4 0.4 % reported NPSV within the last 12 months. This is however, twice the rate of NFHS 3. In contrast, reported IPSV decreased from 2.7 % to 1.2 %. Also, IPSV ever (reported in Appendix 4 and 5) suggests that intimate partner sexual violence overall is less common in 2015/16 than before.

It could indeed be the case that the incidence of NPSV increased as it appears in the media. On the other hand, it is also likely that NPSV was reported more freely in 2015/16 than in 2005/06. The nationwide outrage after several rape cases in India15 _{might have risen awareness for how important it}

is to speak up and that having experienced (sexual) violence is nothing to be ashamed of. However, reported incidents in both rounds seem low and underreporting is very likely. It should be noticed that in NFHS 4 433 women (>0.6 %) refused to answer the questions about sexual violence within the last 12 months, while 176 women (>0.2 %) refused in NFHS 3. Those non-response rates seem surprisingly low given the sensitive questions that have been asked and that the interview took place within the respondent’s home.

Table 3 compares the reported violence variables by the sanitation types outside and non-outside defecation between the two rounds. The mean of women who report NPSV barely differs between the sanitation types in NFHS 4. In NFHS 3 almost twice the share of women who practice outside defection reports NPSV. This indicates a higher correlation between the sanitation type and NPSV in 2005/06 than in 2015/16. IPSV within 12 months prior to the interview do not differ significantly between outside and inside defecation.

Table 3: NPSV and IPSV by Sanitation type, NFHS 4 vs. NFHS 3

Note: Table is based on weighted samples. All numbers are rounded up to four decimal places. *IPSV within 12 months on married sample only

15 _{See footnote 7 for examples.}

NFHS 4: 2015/2016 NFHS 3: 2005/2006 Outside defecation Non-Outside defecation Total Outside defecation Non-Outside defecation Total

As apparent from Figure 1, the prevalence of outside and open defecation differs substantially between rural and urban areas. Figure 3 demonstrates that the area of residence is also relevant for the association between NPSV and sanitation type. In rural, the share of women reporting NPSV is higher when the household sanitation type is outside defecation. The sanitation type seems less relevant in urban. While the share of NPSV when practicing outside defecation exceeded the share of NPSV when reporting non-outside defecation in NFHS 3, the reported share of NPSV is higher for women who not practice outside defecation in NFHS 4.

This fuels the claim that the correlation between NPSV and outside defecation is higher in rural areas than in urban. This is a first indicator that leaving the household for defecation might indeed be a driver of NPSV in the rural setting.

Figure 3: NPSV comparison NFHS 4 and NFHS 3, by outside defecation and area

Note: Figure is based on weighted samples.

0.2 0.4 0.6 Pe rc en ta ge o f w ome n r ep or tin g NP SV in la st 1 2 mo nt hs Rural Urban NFHS 3 NFHS 4 NFHS 3 NFHS 4

5. Empirical analysis

I weighted all regressions by the national domestic violence weight that the DHS assigned to every DV module observation. As the number of eligible women varies across households and only one woman per household was selected for the DV module, the probability for being selected is not the same for every woman. Moreover, there might be a possible lack of independence between women interviewed within the same cluster as those are exposed to the same environment. The complex survey design with stratification and clustering as well as unequal weighing are incorporated to ensure actual national representativeness of result.16 _{Table 3 shows the regression results for the dependent variable}

NPSV for the NFHS 4 data. All results of the logistic regressions are presented in odds ratios (or).

Table 3: The effect of Outside Defecation on NPSV, odds ratios; NFHS 4

(1) (2) (3) Dependent variable NPSV Full sample Urban sample Rural sample Leaving household for defecation 1.097 0.858 1.954**

(0.116) (0.471) (0.533) Age group 20-29 (Reference 15-19) 0.239** 0.0416*** 1.175

(0.152) (0.0278) (0.430)

Age group 30-39 0.249* 0.0591*** 1.043

(0.179) (0.0467) (0.417)

Age group 40-49 0.207* 0.0417*** 1.217

(0.190) (0.0473) (0.587) Primary Education (Reference: no education) 1.056 1.013 1.302

(0.325) (0.792) (0.373) Secondary Education 1.293 2.551 1.169 (0.324) (1.535) (0.320) Higher Education 1.203 2.625 0.843 (0.743) (2.577) (0.410) Currently employed 2.220*** 2.662*** 2.060*** (0.398) (0.731) (0.452) Married (Reference: never married) 1.427 4.159** 0.453**

(0.901) (2.865) (0.169)

Widowed 2.477 9.741** 0.693

(2.056) (10.94) (0.441) Divorced or separated 3.277* 15.48*** 1.023

(2.275) (16.23) (0.679) Number of women in household 1.016 1.053 0.964

(0.136) (0.162) (0.124) Muslim (Reference: Hindu) 0.630 0.534 0.752

(0.384) (0.458) (0.304) Christian 1.008 0.746 1.567 (0.372) (0.442) (0.830) Other/No religion 1.408 0.594 2.053 (0.660) (0.533) (1.093) Scheduled caste/tribe 0.646* 0.498* 0.956 (0.153) (0.191) (0.358) Other Backward Class 0.345*** 0.160*** 0.900

(0.116) (0.0789) (0.373) Urban (Reference: Rural) 2.199**

(0.823)

West (Reference: North-east) 1.286 2.722 0.764 (0.797) (2.636) (0.507) South 9.233*** 28.05*** 3.236** (4.610) (22.33) (1.626) North 0.868 1.525 0.733 (0.372) (1.183) (0.385) North-central 1.936* 3.868* 1.200 (0.757) (2.947) (0.610) East 1.064 0.0946** 0.916 (0.458) (0.111) (0.495) Household has: electricity 0.386*** 0.306 0.524**

(0.118) (0.291) (0.165)

Rudimentary Roofing 0.679 0.194 1.009

(Reference: Natural Roofing) (0.298) (0.242) (0.461)

Permanent Roofing 1.025 2.915 0.866

(0.341) (2.514) (0.298)

Constant 0.00627*** 0.00244*** 0.00314***

(0.00356) (0.00365) (0.00207)

Observations 74,587 22,184 52,403

Notes: Table is on weighted samples, odds ratios in comparison to reference group reported with standard errors in parentheses; Significance levels: *** p<0.01, ** p<0.05, * p<0.1

The logistic probability model results in table 3 show no significant association between outside defecation on NPSV in the full sample. Due to the heterogenous Indian setting, I tested the association between outside defecation and NPSV in a rural and an urban subsample separately.17 _The

results for the rural subsample confirm a significant (at the 5% level) association between the household sanitation facility and NPSV. Women in rural who have to go outside for defecation have almost double (or = 1.954) the risk of NPSV than rural women who have access to private household sanitation. The urban sample does not show any significant association between outside defecation and NPSV.

In the full and the urban sample, all age groups show significantly reduced odds for experiencing NPSV in comparison to the reference group of 15-19-year-olds. Women who are divorced or separated have about three times the odds of experiencing NPSV in comparison to never married women in the full sample, and more than 15 times the odds in the urban sample. Women belonging to other backward classes show significantly lower odds for experiencing NPSV than women who do not belong to those classifications.18 _{Women living in urban have 2.119 the risk of}

NPSV (p<0.05) than women living in rural areas. However, this does not imply that toilet facilities affect the relationship. Surprisingly, in all three samples, women living in the Southern region of India display a much higher risk of NPSV than women living in the North-East of India. Another important driver for NPSV seems to be current employment. Being employed means significantly increased odds in all three samples (p<0.01).

Table 4 shows the results of the logistic regression with the NFHS 3 data. Outside defecation is significantly influencing NPSV in the full and the rural sample. Women without household sanitation facilities have 1.884 times, so almost twice, the risk of NPSV (p<0.1) compared to women who have access to household facilities. While there is no significant effect in the urban sample, women in rural without household sanitation facilities have even 2.891, so almost three times the risk of experiencing NPSV (p<0.05). In contrast to the NFHS 4 results, the odds of experiencing NPSV are significantly higher for a Moslem woman than a Hindu woman, also belonging to a scheduled class or caste means higher odds.

To rule out that omitted variables are accountable for toilet-lacking women’s higher risks of both, experiencing sexual violence and living in households without toilets, I re-run the regressions with IPSV as the dependent variable. The regression results for both NFHS rounds can be found in Appendix 6 and 7. Outside defecation is non-significant in all samples, also in the only rural ones. This disconfirms a simultaneous selection of women into open defecation and sexual violence.

Table 4: The effect of Outside Defecation on NPSV, odds ratios; NFHS 3

(1) (2) (3) Dependent variable NPSV Full sample Urban sample Rural sample Leaving household for defecation 1.884* 0.712 2.891**

(0.652) (0.334) (1.363) Age group 20-29 (Reference 15-19) 1.641 3.013 1.491

(0.638) (2.353) (0.720)

Age group 30-39 0.535 0.851 0.508 (0.222) (0.736) (0.256)

Age group 40-49 0.340** 0.009*** 0.419

(0.183) (0.0123) (0.246) Primary Education (Reference: no education) 0.778 0.232* 1.032

(0.285) (0.175) (0.419) Secondary Education 0.757 0.221** 1.080 (0.263) (0.153) (0.425) Higher Education 1.138 0.274 1.916 (0.942) (0.268) (2.351) Currently employed 1.237 1.207 1.263 (0.393) (0.643) (0.493) Married (Reference: never married) 0.388** 0.272** 0.393**

(0.165) (0.169) (0.181)

Widowed 2.044 2.779*

(1.098) (1.658)

Number of women in household 1.155 0.676 1.313 (0.255) (0.220) (0.281) Muslim (Reference: Hindu) 3.770*** 2.945 4.922***

(1.632) (1.944) (2.902) Christian 0.896 0.375 1.275 (0.595) (0.407) (0.964) Other/No religion 0.147*** 0.158** 0.182*** (0.0716) (0.125) (0.112) Scheduled caste/tribe 3.079*** 3.609** 3.041* (1.320) (1.897) (1.848)

Other Backward Class 1.862 2.139 1.837

(0.796) (1.117) (1.129) Urban (Reference: Rural) 0.738

(0.259)

West (Reference: North-east) 0.363 0.288 0.402 (0.230) (0.300) (0.297) South 0.256** 0.0395*** 0.392 (0.141) (0.0333) (0.244) North 0.353** 0.755 0.114*** (0.171) (0.566) (0.0841) North-central 0.309** 0.0275*** 0.400 (0.154) (0.0317) (0.228) East 0.321*** 0.429 0.282** (0.141) (0.311) (0.149) Household has: electricity 0.822 1.094 0.731

(0.284) (0.571) (0.271)

Rudimentary Roofing 1.257 0.613 1.322

(Reference: Natural Roofing) (0.802) (0.500) (1.147)

Permanent Roofing 2.107* 0.351* 2.722**

(0.824) (0.193) (1.253)

Constant 0.00116*** 0.0410* 0.000387***

(0.00128) (0.0672) (0.000524)

Observations 76,611 33,284 42,302

5.2 Alternative estimations

I carried out several alternative estimations. First, I used an alternative method, the probit regression, which, as anticipated yielded the same results in terms of significance and direction of effects (results not shown).19

Then, I focused on the more recent data from 2015/16 and performed alternative estimations on the rural sample only. I check whether the significant results for the rural sample are sensitive to specifications of the model. Regression results are shown in Appendix 8. Therefore, I leave out current employment in regression [1]. The odds for experiencing NPSV when not having a household facility (or = 2.081, p-value<0.01) is larger than if the control is included (or = 1.954, p-value<0.05), and is then significant at the one percent level. In regression [2] I use an alternative to control for prosperity of the household. Instead of controlling for household assets (as done before and by Jadhav et al. (2016)), I now control for the rural specific wealth index. I did not apply the wealth index before due to its composition. Including the rural wealth index decreased the effect of outside defecation which I expected as the toilet facility is already included in the wealth index. Overall however, it barely changes the results (or = 1.860, p-value<0.05).

Also, I tested different measures for the lack of sanitation. First, open defecation has been coded 1 if “no facility/bush” was selected and 0 if any type of toilet facility is used. Regression [3] shows that the odds ratio of 1.779 remains significant, however only at the 10 % level. Second, unimproved sanitation has been coded 1 if the facility did not meet the WHO requirements for an improved sanitation and 0 if the facility is considered improved.20 _{Also, with this specification the odds ratio of}

1.681 remains significant at the 10% level (Regression [4]).

Finally, I carry out the regression with outside defecation on the unweighted rural sample in regression [5]. Outside defecation remains significantly correlated with NPSV in the rural sample at the ten percent level when all observations of the selected rural women are weighted equally, and robust standard errors are applied. In this case a woman living in a household with several other eligible women does not have a higher representative weight than a woman who is the only one in her household. The sampling weights adjust for differences in the probability for being selected for the interview and for non-response to relevant questions and I therefore prefer the weighted estimations.

5.3 Discussion

In summary, the logistic regression results confirm a significant association between NPSV and the lack of private household sanitation in the rural samples of NFHS 3 and NFHS 4. Those results support anecdotal evidence and the results of qualitative studies in rural India. It thus confirms the hypothesis that rural women without private household sanitation have a higher risk to experience NPSV than women with private facilities. The heighted fear of violence when defecating outside that rural women report (inter alia in Sahoo et al. 2015) is quantified by a higher risk for actually experiencing NPSV in rural India.

However, no significant link can be found in the full sample of 2015/16 while the 2005/06 data suggests an effect of outside defecation on NPSV at the 10% level. Also, the odds ratios in rural are lower for NFHS 4 than for NFHS 3. This is surprising as I did not expect those effects to be period specific as such. Accordingly, the significance and magnitude of the association between experiencing NPSV and access to a private household facility seems to vary depending on the period, sample and model specification.

The latter became apparent through the alternative estimations. Leaving out a control variable such as current employment significantly increases the significance to the 1 % level (Appendix 8, Regression [1]). When the measure for lack of sanitation is restricted to open defecation only and women who use shared facilities are not included, the significance of the effect is attenuated to the 10 % level (Regression [3]). When extending the measure of lacking sanitation from outside defecation to unimproved sanitation, attenuated results are less surprising. The toilet facilities that are related to the women are based on the facility that has been reported in the household survey. This does not necessarily mean that all household members consistently use this facility. WASH research in the Indian context suggests that – beside the pure access to sanitation – also certain characteristics of a toilet facilities matter for whether a facility is actually used (See Section 3, Coffey et al. 2017a). Relevant characteristics are cleanliness and functionality of the facility, a nearby handwashing possibility or whether toilet use is possible undisturbed. Unfortunately, this cannot be measured with the available DHS data.21 _{However, I tried to incorporate that Indian women might also base the}

decision where to go for defecation on what kind of flush is available and how feces are managed. The results show that women with an unimproved facility have higher odds of experiencing NPSV than those using an improved facility, however only significant at the 10 % level. However, it is still not feasible to distinguish whether women actually use the reported facility and the significance of

unimproved sanitation might be driven by the large prevalence of outside and open defecation that is included in the unimproved sanitation measure.

The missing significant effect in the NFHS4 full sample in comparison to the one of NFHS 3 point in the direction that results are indeed period specific, unlike assumed in Section 3 and claimed by Jadhav et al. 2016. One possible explanation might be that the shared facilities – which I included in the outside defecation measure – are better connected to the village, situated in less remote places or equipped with lighted paths which might obstruct violent attacks. Increased awareness of the risks due to e.g. better media reporting might have also resulted that women do not walk to open defecation sites or public toilets alone and the whole community might be more alert. Thus, outside sanitation spots in general might be safer in 2015/16 than in 2005/6. The share of outside defecation is about 20 % smaller in NFHS 4 than in NFHS 3. Latrine construction and sanitation efforts by the GoI during that period might have mitigated the risk for sanitation-related NPSV to some extent.

Therefore, it seems that outside defecation is less relevant for NPSV in 2015/16 than it used to be. Also, the included control variables indicate that NPSV in 2015/16 is correlated with other factors than ten years before. Current employment is highly significantly correlated with NPSV in the NFHS 4 samples and non-significant in the NFHS 3 round.

As already mentioned, it is surprising that a woman living in South India have a much higher risk of NPSV than a woman in North-East. All the other states except of North Central have lower odds than the North East (mostly non-significant). When controlling for state instead of region dummies, the only significant Southern state is Karnataka (regression results not shown). Further investigating the data shows that 40 of the 193 reported NPSV cases are from this state.22 _This

association might be rather driven by factors that lie outside the range of those data sets. The exorbitantly high correlation between NPSV and South seems moreover not be rooted in a lack of sanitation facilities as a separate regression on the South sample confirmed (outside defecation is insignificant, results not shown). Nevertheless, one possible explanation could be a different attitude towards reporting VAW in Karnataka in comparison to other Indian states.

Next, I will discuss possible endogeneity issues in the model’s specification. Generally, caution is required when trying to establish a causal relationship between information that has been collected at the same point of time as it is the case with the cross-sectional data sets used for this paper. Thus,

the DHS data does not allow to observe the direct link between going outside for defecation and women’s experience of NPSV as one cannot determine the timely ordering of exposure and outcome. As mentioned above, this is one of the reasons why the violence measure is restricted to incidents within the last 12 months. Accordingly, it could be the case that outside defecation is not exogenous but correlated with the error term. Omitted variables as the cause for endogeneity are relatively likely in cross-sectional analysis. For the DHS many different variables from various areas are collected to manifest the situation in demographics and health. Thereby, it does not capture the circumstances of certain incidents or how a disease was contracted, nor the motives for e.g. practicing open or outside defecation. However, I assume that outside defecation is exogenous to the model, especially in the Indian context, where it is inter alia driven by tradition and cultural factors. Many potential risk factors for NPSV (such as individual ones) are simply not available. Yet, I deem it improbably that those also affect outside defecation. It is possible that the causal relationship is the other way around and NPSV influences the probability that women practice outside defecation. Women might refrain from using certain open defecation spots because of the fear of NPSV. Tragically, it is impossible to opt against outside defecation if no household facility is available. Thus, outside defecation is rather a choice without alternatives and reversed causality unlikely.

The design of the DV module also implies that the measure of NPSV is imperfect.23 _{I cannot}

rule out that reported NPSV is still sexual family violence. Also, there is no information where the rape took place. Further, NPSV only captures reported forced sexual intercourse. Other attacks such as rape attempts or other sexual assault are not captured in the measure. Also, not only physical attacks are considered harassment, but also starring, obscene comments and gestures and insults which are also entirely missing from the measure. Moreover, the measure of NPSV only captures reported incidence of the last 12 months of girls and women from 15 to 49 years of age. However, qualitative surveys (such as Nallari 2015) report that especially adolescent girls fear rape and sexual assault and the anecdotal evidence from the media shows that also substantially younger girls are affected. This suggests that the real numbers of NPSV are a lot higher than the NFHS data suggests.

Nevertheless, NPSV is – to my perception – the best feasible measure in this setting. The DV module aims primary on quantifying domestic violence and applying it for non-domestic violence means to compromise. Thus, this cross-sectional data set and the specification it allows are not perfect for quantifying sexual violence by non-family members during the use of an outside sanitation facility, However, in absence of an India-wide crime victimization study, the DHS data is the best available option to quantify the effect with a large data set and without collecting own data in field interviews.

6. Conclusion

In this paper, I analyzed the link between access to household sanitation and women’s risk of experiencing NPSV. By utilizing a large secondary data set, I find a significantly increased risk of NPSV for women who practice outside defecation in rural areas.

It is important to note that the lack of toilet facilities is not considered the cause of rape or sexual violence. However, not having a safe place for toilet use facilitates the violent act. Despite several limitations, which can mainly be traced back to the design of the underlying data set, the results have important implications for sanitation planning. They particularly demonstrate that solving sanitation insecurity24 _{is as important as providing sanitation access. It is of crucial importance that}

the voices of women are heard in the process of sanitation planning, in order to ensure that location and design of the facility provide a maximum of sanitation security.

Despite the need for gender-responsive sanitation solutions due to dignity, privacy, safety and health-reasons, emphasizing gender-related benefits in sanitation campaigns remains problematic (Routray et al. 2017). Instrumentalization of women’s benefits as a motivator in campaigns could reinforce the perception that improved sanitation only matters for women, rather than for the whole community. Also, patriarchal campaigns present a lack of household sanitation as only shameful for women.25 _{This fails to communicate the necessity that also men have to change their defecation}

behavior. Also, gender-based power dynamics at the household level needs to be addressed as men are traditionally the economic decision maker within the households (Coffey et al. 2017a, Khanna and Das 2016).

Consequently, sanitation campaigns in India must offer gender-responsive sanitation solutions, address people’s real concerns about pit emptying and pollution and incorporate the demand-side drivers of household’s sanitation choice. Equally important, they must transmit the message that especially the practice of open defecation is a collective action problem with persisting contaminating consequences for the whole community until everybody stops.

In India, more than anecdotes are needed to promote policy change and to address VAW at its roots. However, the currently available data sets, which are nation-wide representative, have important

24 _{Sanitation insecurity is defined as “insufficient and uncertain access to a socio-cultural and social environments that}

respect and respond to the sanitation needs of individuals, and to adequate physical spaces and resources for independently, comfortably, safely, hygienically, and privately urinating, defecating, and managing menses with dignity at any time of day or year as needs arise in a manner that prevents fecal contamination of the environment and promotes health” (Caruso et al. 2017, p. 9).

25_{“Daughters and Daughters-in-law shouldn’t go outside, build a toilet inside your house” is one of the slogans that}

References

Barnard, S., Routray, P., Majorin, F., Peletz, R., Boisson, S., Sinha, A., & Clasen, T. (2013). Impact of Indian Total Sanitation Campaign on latrine coverage and use: a cross-sectional study in Orissa three years following

programme implementation. PloS one, 8(8), e71438.

BBC (2017). Bangalore New Year: 'People were grabbing, groping';

https://www.bbc.com/news/world-asia-india-38504186, retrieved June 3rd _2019.

BBC (2014). Why India's sanitation crisis kills women;

https://www.bbc.com/news/world-asia-india-27635363, retrieved June 4th _2019.

BBC (2014b). India gang rapes: Police sacked over hanged girls

https://www.bbc.com/news/world-asia-india-27639735, retrieved June 4th ₂₀₁₉

BBC (2013). India Bihar rapes 'caused by lack of toilets';

https://www.bbc.com/news/world-asia-india-22460871, retrieved June 4th _2019.

Bhattacharya, P. and Kundu, T. (2018). 99% cases of sexual assaults go unreported, govt data shows;

https://www.livemint.com/Politics/AV3sIKoEBAGZozALMX8THK/99-cases-of-sexual-assaults-go-unreported-govt-data-shows.html, retrieved June 10th _2019.

Caruso, B. A., Clasen, T. F., Hadley, C., Yount, K. M., Haardörfer, R., Rout, M.& Cooper, H. L. (2017). Understanding and defining sanitation insecurity: women’s gendered experiences of urination, defecation and menstruation in rural Odisha, India. BMJ global health, 2(4), e000414.

Census India (2011): Availability and Type of latrine facility;

http://censusindia.gov.in/2011census/hlo/Data_sheet/India/Latrine.pdf

Coffey, D., Spears, D., & Vyas, S. (2017a). Switching to sanitation: understanding latrine adoption in a representative panel of rural Indian households. Social Science & Medicine, 188, 41-50.

Coffey, D., Gupta, A., Hathi, P., Spears, D., Srivastav, N., & Vyas, S. (2017b). Understanding open defecation in rural India: Untouchability, pollution, and latrine pits. Economic and Political Weekly, 52(1), 59-66.

CNN (2018). Six years after Delhi bus attack, India rape crisis shows no sign of slowing;

https://edition.cnn.com/2018/12/22/asia/india-rape-crisis-delhi-intl/index.html, retrieved June 4th 2019. Dickinson, K., & Pattanayak, S. K. (2009). Open sky latrines: Do social interactions influence decisions to use toilets. Working paper.

Frye, V., & O’Campo, P. (2011). Neighborhood effects and intimate partner and sexual violence: Latest results. Journal of urban health, 88(2), 187-190.

Gonsalves, G. S., Kaplan, E. H., & Paltiel, A. D. (2015). Reducing sexual violence by increasing the supply of toilets in Khayelitsha, South Africa: a mathematical model. PLoS one, 10(4), e0122244.

Gulfnews (2014). 'Boys will be boys': Outrageous rape comments by India politicians;

https://gulfnews.com/world/asia/india/boys-will-be-boys-outrageous-rape-comments-byindia politicians-1.1347378, retrieved May 10th _2019.

Hirve, S., Lele, P., Sundaram, N., Chavan, U., Weiss, M., Steinmann, P., & Juvekar, S. (2015). Psychosocial stress associated with sanitation practices: experiences of women in a rural community in India. Journal of Water,

Sanitation and Hygiene for Development, 5(1), 115-126.

Hueso, A., & Bell, B. (2013). An untold story of policy failure: the Total Sanitation Campaign in India. Water

Human Rights Watch (2014): Cleaning Human Waste - "Manual Scavenging," Caste, and Discrimination in India https://www.hrw.org/report/2014/08/25/cleaning-human-waste/manual-scavenging-caste-and-discrimination-india, retrieved May 19th 2019

Jadhav, A., Weitzman, A., & Smith-Greenaway, E. (2016). Household sanitation facilities and women’s risk of non-partner sexual violence in India. BMC public health, 16(1), 1139.

Karpagam S. in idronline (2019): The problems with the National Family Health Survey; https://idronline.org/the-problems-with-the-national-family-health-survey-nfhs/

Khanna, T., & Das, M. (2016). Why gender matters in the solution towards safe sanitation? Reflections from rural India. Global public health, 11(10), 1185-1201.

Kiss, L., Schraiber, L. B., Hossain, M., Watts, C., & Zimmerman, C. (2015). The link between community-based violence and intimate partner violence: The effect of crime and male aggression on intimate partner violence against women. Prevention science, 16(6), 881-889.

Nallari, A. (2015). “All we want are toilets inside our homes!” The critical role of sanitation in the lives of urban poor adolescent girls in Bengaluru, India. Environment and Urbanization, 27(1), 73-88.

Parks, M. J. (2014). Urban poverty traps: neighbourhoods and violent victimisation and offending in Nairobi, Kenya. Urban Studies, 51(9), 1812-1832.

Routray, P., Schmidt, W. P., Boisson, S., Clasen, T., & Jenkins, M. W. (2015). Socio-cultural and behavioural factors constraining latrine adoption in rural coastal Odisha: an exploratory qualitative study. BMC public

health, 15(1), 880.

Sahoo, K. C., Hulland, K. R., Caruso, B. A., Swain, R., Freeman, M. C., Panigrahi, P., & Dreibelbis, R. (2015). Sanitation-related psychosocial stress: a grounded theory study of women across the life-course in Odisha, India. Social science & medicine, 139, 80-89.

Srivastav and Gupta (2015). Why Using Patriarchal Messaging to Promote Toilets is a Bad Idea;

https://thewire.in/culture/why-using-patriarchal-messaging-to-promote-toilets-is-a-bad-idea, retrieved June 16th

2019.

The Guardian (2014). Two girls died looking for a toilet. This should make us angry, not embarrassed; https://www.theguardian.com/global-development/2014/jun/01/girls-toilet-rape-murder-anger-embarrassment, retrieved June 5th _2019.

Truelove, Y. (2011). (Re-) Conceptualizing water inequality in Delhi, India through a feminist political ecology framework. Geoforum, 42(2), 143-152.

UNICEF (2013). Sanitation Factsheet for World Toilet Day 2013;

https://www.unicef.org/toilets4all/WTD_Sanitation_Factsheet_November_2013.pdf.

United Nations (2015). The World's Women 2015 Trends and Statistics. New York: United Nations, Department of Economic and Social Affairs, Statistics Division. Sales No. E.15.XVII.8.

Winter, S. C., & Barchi, F. (2016). Access to sanitation and violence against women: evidence from Demographic Health Survey (DHS) data in Kenya. International journal of environmental health research, 26(3), 291-305. WHO/UNICEF (2017). Progress on Drinking Water, Sanitation and Hygiene, 2017 Update and SDG Baselines Geneva.

WHO/UNICEF (2013). Ending Preventable Child Deaths from Pneumonia and Diarrhoea by 2025.

Appendix

Appendix 1: Full, Rural and Urban sample by States, 2015/2016

State/union territory Frequency (%) Full Sample Frequency (%) Urban Sample Frequency (%) Rural Sample North

Chandigarh* 83 (0.11) 74 (0.33) 9 (0.02)

Delhi* 450 (0.60) 442 (1.98) 8 (0.02)

Haryana 2,149 (2.86) 777 (3.49) 1,372 (2.60) Himachal Pradesh* 1,771 (2.36) 163 (0.73) 1,608 (3.05) Jammu and Kashmir* 4,274 (5.70) 939 (4.22) 3,335 (6.32) Punjab* 1,986 (2.65) 750 (3.37) 1,236 (2.34) Rajasthan* 4,009 (5.34) 1,059 (4.75) 2,950 (5.59) Uttarakhand 1,636 (2.18) 473 (2.12) 1,163 (2.21) North-Central Chhattisgarh* 2,449 (3.26) 671 (3.01) 1,778 (3.37) Madhya Pradesh 5,925 (7.90) 1,767 (7.93) 4,158 (7.88) Uttar Pradesh* 8,658 (11.54) 2,416 (10.85) 6,242 (11.84) East Bihar 4,490 (5.99) 681 (3.06) 3,809 (7.22) Jharkhand* 2,973 (3.96) 727 (3.26) 2,246 (4.26) Odisha* 3,662 (4.88) 756 (3.39) 2,906 (5.51) West Bengal 1,850 (2.47) 487 (2.19) 1,363 (2.58) North-East Arunachal Pradesh* 1,524 (2.03) 302 (1.36) 1,222 (2.32) Assam 3,027 (4.04) 450 (2.02) 2,577 (4.89) Manipur 1,391 (1.85) 491 (2.20) 900 (1.71) Meghalaya 918 (1.22) 200 (0.90) 718 (1.36) Mizoram* 1,272 (1.70) 637 (2.86) 635 (1.20) Nagaland* 1,102 (1.47) 372 (1.67) 730 (1.38) Sikkim 624 (0.83) 178 (0.80) 446 (0.85) Tripura 675 (0.90) 183 (0.82) 492 (0.93) West

Dadra and Nagar Haveli* 99 (0.13) 45 (0.20) 54 (0.10) Daman and Diu* 220 (0.29) 131 (0.59) 89 (0.17)

Goa 559 (0.75) 282 (1.27) 277 (0.53)

Gujarat* 3,642 (4.85) 1,371 (6.15) 2,271 (4.31) Maharashtra 2,885 (3.85) 1,070 (4.80) 1,815 (3.44)

South

Andaman and Nicobar

Islands 303 (0.40) 59 (0.26) 244 (0.46) Andhra Pradesh 1,086 (1.45) 352 (1.58) 734 (1.39) Karnataka 2,449 (3.26) 839 (3.77) 1,610 (3.05) Kerala* 1,623 (2.16) 654 (2.94) 969 (1.84) Lakshadweep* 112 (0.15) 96 (0.43) 16 (0.03) Puducherry 517 (0.69) 390 (1.75) 127 (0.24) Tamil Nadu 3,783 (5.04) 1,684 (7.56) 2,099 (3.98) Telangana 842 (1.12) 309 (1.39) 533 (1.01) Total 75,018 (100) 22,277 (100) 52,741 (100)

Appendix 2: Full, Rural and Urban sample by States, 2005/2006

State/union territory Frequency (%) Full Sample Frequency (%) Urban Sample Frequency (%) Rural Sample North

Appendix 3: Prevalence of open defecation, comparison NFHS 3 and NFHS 4

Note: Graph is based on weighted samples; Regions are divided as follows: North-east: Arunachal Pradesh, Assam, Manipur, Meghalaya, Mizoram, Nagaland, Sikkim, Tripura; West: Goa, Gujarat, Maharashtra; South: Andhra Pradesh, Karnataka, Kerala, Tamil Nadu; North: Haryana, Himachal Pradesh, Jammu and Kashmir, Punjab, Rajasthan, Uttarakhand, Delhi; North-central: Chhattisgarh, Madhya Pradesh, Uttar Pradesh; East: Bihar, Jharkhand, Odisha, West Bengal; Delhi is only union territory in NFHS 3.

0 20 40 60 80 100 O pe n de fe ca tio n %

North-East West South North North-Central East

NFHS 3 NFHS 4 NFHS 3 NFHS 4 NFHS 3 NFHS 4 NFHS 3 NFHS 4 NFHS 3 NFHS 4 NFHS 3 NFHS 4

Appendix 4: Descriptive statistics: Full, Urban, Rural Sample of Women 2015/2016

Variables Full Sample

N= 75,018 Urban Sample N=22,277 Rural Sample N=52,741 Violence

Non-Partner Sexual Violence within last

12 months 0.004 (0.061) 0.006 (0.076) 0.003 (0.052) Intimate-Partner Sexual Violence within

last 12 months 0.011 (0.106) 0.012 (0.108) 0.011 (0.105) Sexual Violence from anyone ever 0.052 (0.222) 0.039 (0.195) 0.058 (0.234) Intimate-Partner Sexual Violence ever 0.068 (0.252) 0.053 (0.225) 0.075 (0.264) Physical violence by family member ever 0.082 (0.275) 0.079 (0.270) 0.084 (0.277)

Sanitation

Open defecation 0.345 (0.475) 0.093 (0.290) 0.473 (0.499) Toilet facility: pit 0.091 (0.287) 0.060 (0.237) 0.106 (0.308) Toilet facility: toilet 0.565 (0.496) 0.848 (0.359) 0.421 (0.494) Outside defecation 0.436 (0.496) 0.233 (0.423) 0.540 (0.498) Using a shared/public facility 0.091 (0.289) 0.139 (0.346) 0.067 (0.250) Private facility if using a facility 0.860 (0.347) 0.846 (0.361) 0.873 (0.333) Shared facility if using a facility 0.140 (0.347) 0.154 (0.361) 0.127 (0.333) Improved sanitation facility 0.607 (0.488) 0.871 (0.335) 0.472 (0.499)

Demographic characteristics Age in years 30.38 (9.788) 30.807 (9.678) 30.083 (9.837) Age 15-19 0.169 (0.374) 0.151 (0.358) 0.177 (0.382) Age 20-29 0.324 (0.468) 0.319 (0.466) 0.327 (0.469) Age 30-39 0.281 (0.449) 0.294 (0.455) 0.274 (0.446) Age 40-49 0.227 (0.419) 0.237 (0.425) 0.221 (0.415) Education in years 6.912 (5.223) 8.892 (5.112) 5.905 (4.985) No education 0.271 (0.444) 0.154 (0.361) 0.330 (0.470) Primary education 0.122 (0.327) 0.095 (0.293) 0.136 (0.343) Secondary education 0.482 (0.500) 0.526 (0.499) 0.459 (0.498) Higher education 0.125 (0.331) 0.225 (0.418) 0.075 (0.263) Currently employed 0.295 (0.456) 0.240 (0.427) 0.323 (0.467) Relationship status Never married 0.250 (0.433) 0.274 (0.446) 0.238 (0.426) Ever married 0.750 (0.433) 0.726 (0.446) 0.762 (0.426) Currently married 0.708 (0.455) 0.678 (0.467) 0.723 (0.447) Widowed 0.031 (0.172) 0.035 (0.184) 0.028 (0.165) Divorced or separated 0.012 (0.108) 0.013 (0.112) 0.011 (0.105) Number of 15-49-year-old women in