 Okay, so good afternoon everyone, my name is Sudeepa and my paper is about how perceived crime of households affects early marriage in India. So as Kunal mentioned, this is also part of this project, Women's Work. Okay, so the objective is, as I mentioned, to look at how perception of crime in the locality where the households live impacts the decision to marry the daughters and as well as the sons of the households because this is particularly because in India more than 80% of marriages are organized arranged by the family. Yeah, so that's why in that context it's very important to look at it. We also look at both gender-neutral and gender-specific crimes in the locality. So we have information on whether there is theft in the locality, robbery, attack or any kind of harassment, gender-specific harassment. So these are reported by the households, so that's what we use. And we also explore the channels through which this relationship may hold. Okay, so just to give you a little bit of background, I think that the importance of early marriage, why it's important, I don't have to motivate it. The keynote speaker also has mentioned it, Kunal has mentioned it in terms of the importance of it, in terms of, in the context of women's labour force participation. So I would just mention, if we look at the incidence of child marriage then around half of the child bride live in South Asia. And when we look at India, despite several efforts made by national and international agencies, early marriage remains perversive. If you look at the statistics, 27% in 2015, 16 were married girls were married before the legal age of marriage, which is 18. Well, it has increased now, but when I did this study it was 18. And when we look at the recent data from 2019 and 21, it's 23%. So it has, the reduction is really about four percentage point. And when we look at age at marriage, I mean, there's a lot of policies across India which are basically giving incentives to the parents so that they keep cash transfer to parents so that they keep their daughters in school and automatically there will be a reduction in age at marriage. But despite all that, when we look at age at marriage in India, it's actually increasing slowly like one year per decade. So that's the progress. So in this context, I think that it's important to look at different factors that are important for age at marriage and child marriage. So looking at the literature, there is a huge literature on child marriage, early marriage, generally focusing on girls, girl child marriage because even though there is child marriage for men as well, but I think the literature has mostly focused on women because I think the consequences are more serious for women. Yeah, so just to summarize, I have just categorized the literature into two broad categories, like one looking at the determinants of early marriage and the other looking at the consequences. So if we look at the determinants, the researchers have highlighted household poverty, parental education, access to opportunities. So these are mostly relevant for early marriage of girls. And when we look at consequences, early and teenage marriage are found to be associated with poor human capital formation because in most of the countries they have found that early marriage leads to school dropout, which that means that there will be poor human capital formation and which in turn affect the labor market outcome of the women. And it's also found to be associated with child health and maternal health because early marriage generally leads to early pregnancies and which has adverse consequence on child health as well as maternal health. So it's also found to be associated with high rates of domestic violence because the girls who are married at a tender age, they have lower agency bargaining power. So that's where it's found to be highly associated with domestic violence. So what is missing in the literature is that most of these papers they have pointed out about the role of safety, safety concern. I think before the pre-lunch session we had a paper where the researcher has mentioned about safety concern and women's labor force participation. So in this literature as well they have mentioned about how women's safety concern and the value the society places on women's stacity may play a role in deciding the age of marriage of women. But if you look at empirical evidence it's not, it's mostly missing. So it's generally a conjecture. So we don't have any empirical studies giving evidence. So that's where I try to pitch in my study. That's where my contribution is. Okay, so these are the research questions that I try to answer is household's perception of crime in the locality associated with early marriage of girls. We estimate this relationship for adolescent girls who were 12 to 16 year in the baseline. We also look at different gender neutral crimes, whether they have any impact. Does it, they are disproportionately hard young women than men of comparable age groups. So we also look at, we take a sample of boys who are 15 to 19 year old so that they are comparable with the girls. What are the, then we look at the mechanisms or channel behind the link between early marriage of girls and crime against women in the locality. Okay, so how do we answer these questions? So we use a panel data from India Human Development Survey. It's a nationally representative survey of 41,000 households and around 215,000 individuals coming from all over India. And it has collected information on the same individuals and households in 2004-5 and 2011-12. And this panel element allows us to look at crime against women in 2004-5 and to look at the outcome variables, observed outcome variables in 2011-12. And we also use, so because of, I should mention this, because of battery local residence system in India, generally what happens when you have a panel survey in the follow-up round, the women who are married, they drop out of the sample because they move to a different households. But in this survey they have a tracking schedule, so through which they have tracked most of the women and they have collected information on their marriages and the migration status where they live and all that. So I use that information. Yeah, so I use information on crime against women in the neighborhood collected from households and use age at marriage to define early marriage or child marriage for adolescent girls aged 12 to 16 at baseline and unmarried. So this is the sample just to summarize. The final sample is 12,000 female and around 11,000 male. And when we look at the, like I said, if you look at the migration, so more females migrated between rounds than males, but around 72% were tracked by the surveyors. Okay, so these are my main variables. The dependent variable is married between round and early marriage. So it's just one, zero, one if marital status is 2012 is married and marital status in 2012 is unmarried equal to zero. We define early marriage as one if age at marriage is less than legal age 18 for girls and 21 for boys and zero if it is greater than equal to legal age or unmarried. So our main explanatory variable is crime against women reported by households at baseline. So the questions they were asked, how often are unmarried girls harassed in your village neighborhood, which is coded as zero for never and one for often or sometimes. So we just, we define it at neighborhood level by looking at the percentage of households who said yes often or sometimes. So our empirical strategy is very simple. We run a linear probability model where married dummy and early marriage dummy is the dependent variable and then we have, we also explored the mechanisms by estimating the relationship. I think that I'll explain that later on. So I'll go to the endogenity. So to tackle that, we try to tackle to some extent the endogenity in the main dependent variable, which is crime against women by aggregating. So it is possible that the households where there are more unmarried girls, they will tend to report this thing, the harassment of, they will perceive the crime against women in the locality higher than those who do not have unmarried girls. So what we do, we aggregate this crime, reported crime at village level excluding the households own perception. And also we control additional household level characteristics to proxy conservative attitude of households. So these are our main equations. Yeah, so, okay, I'll just keep this. So this is the summary statistics. I will just summarize a few points. So we have, like I said that we have a tracking data that I have used. So we don't have information for the outcome variables for around 20% of the female sample. So I'll talk about that later. And around 47% of female in the sample were married between rounds, 15% had early marriage and around 12.6% of households reported harassment of unmarried girls in the locality in 2004-2005 and it actually increased in 2011-20%. So this is the main regression result and you can see that the gender-specific crime which is harassment of unmarried girls is significantly positively associated with marriage decision between rounds as well as early marriage. And all of the crime rates remain insignificant, which are basically gender-neutral crimes. We also do it for, in this regression, we include the male sample. Yeah, and so we have an interaction term. So this is on pooled sample and we find households' perception of crime, both gender-neutral and gender-specific, have no significant association with the marriage decision of boys. So we explode the channels through which this relationship may hold. And the first thing we look at is the stigma cost attached to the victim of sexual harassment and it could be that the cost is higher in the marriage market. So this is one paper which is highlighted. This is a cross-country study of mate preferences and it has highlighted that Indian men put more weightage on their spouses' sexual priority at marriage compared to their physical appearance and this is compared to men from the rest of the world. So in this context, I think that stigma cost would be higher in the marriage market and also in conservative society. So what we do, we conduct our analysis on two sub-samples, so two sets of households. So one set of households where this conservative practice of Pardha or Gungar, this is a practice where women cover their face with a piece of cloth from outsider, generally from men. So we have that information whether the households practice this conservative ritual or not. So we use that as a proxy to separate out households that are conservative and the households which do not have this practice. And we run our regression for both sets of households. And we also look at northern region and southern region separately because in the literature on gender inequality in India, it's established that northern regions are more patriarchal. So southern region, the states in southern region are more gender egalitarian, so this is already established. So that's why we conduct our analysis separately for northern region and southern region. And we also look at concern for well-being. So it could be that the households are also concerned about the well-being of their daughters. So that's why they try to marry them off to a different neighborhood than their own where they think that there's more crime. So that's what we do. We explore by investigating whether households with higher perception of crime marry their daughters off outside the village or town. So just to give you the main result from the first channel that you can see that we find, I'll use this, yeah. So you can see that we find that this crime against women is significant for the conservative households practicing Pardha and for the households living in northern region. And we do not find any significant association for the other sets of households. Coming to the second channel, we see, so we done another regression to look at whether the households who perceive that their locality is not safe, they are marrying their daughters off to a different village. And that's what we find that crime against women is significant, significantly associated for marriage outside the village. Yeah, so we have also done some robustness check and the results remain the same, so I don't have enough time. So I'll just go to the sample attrition. So as I mentioned that we have around 20% for a 20% sample, we do not have the information about marriage. So we try to look at this by estimating an attrition equation where sample attrition is the dependent variable. So it could be basically what we try that whether these sample dropouts are non-random and are from high crime or low crime localities as compared to the sample included in the analysis, so that's what we try to test. So we just run an attrition equation where sample attrition is the dependent variable and we regress it on all the explanatory factors that are included in our main regression. And the results show that none of the crime variables is significant, so we just take it that probably it's unlikely that there is selection based on crime. Yeah, so I will just conclude. So what we find, we find that households' perception about crime against women significantly increases the likelihood of marriage and early marriage of young girls and we do not find any link between perceived crime in the locality and the marriage decision of young boys. And we also do not find any link between perception of other gender neutral crimes such as robbery and theft. So when we look at the mechanisms, it suggests that this relationship depends on the extent to which the conservative gender norms are practiced and female chastity is valued in the society. OK, so I'll just stop here.