 one and all present here. First of all, I would like to thank UNU wider for providing me this opportunity to present my work in this conference. The topic of my presentation is health shocks and intergenerational transmission of inequality. This is an empirical study based on the southern state of India called Andhra Pradesh. This will be the brief outline of my talk. Income shocks, the effect of income shocks on households in developing countries has been analyzed by a wide strand development economics literature. Now these shocks can be covariate in nature affecting all households in a particular community like droughts and floods, or they can be idiosyncratic in nature affecting that are specific only to individual households. Now in my study, I deal with the welfare impact of a particular type of idiosyncratic income shocks called the health shocks. Health shocks are most common idiosyncratic shocks and also the most important reasons for this distance of households into poverty in developing countries. Now why is this so? This is because health shocks entail economic costs to the households. These costs can be direct costs in terms of medical expenditure or indirect costs in terms of loss of productive labor time and thereby labor earnings. So in order to cope with the burden of these economic costs, households use a variety of formal and informal mechanisms like savings, transfers, credit sale of assets, taking extra work and so on. But whether the households are actually able to cope with the burden of these costs depends upon a variety of factors. These include our own resources possessed by the household like human capital, physical capital, social capital, financial capital and so on. Thus we might, one might expect that the poor of households in developing countries might be doubly disadvantaged to cope with the burden of health shocks. This is because they neither have access to own resources nor they have access to credit, well-developed credit and insurance markets. Now, hence these households might adopt costly strategies like withdrawing children from school, which in turn have implications for vulnerability to future shocks and intergenerational transmission of inequality and poverty. So in my study, I deal with the effect of parental health shocks on investments in children's human capital. This study will also throw some light on important dimensions like role of timing of the shocks, pathways through grease-free shocks affect human capital investment, the differential effects of maternal and paternal shocks and their differential effects on younger and older children and the importance of school quality. Now, coming to the theoretical background for this study, the main study, so this, the theory framework is drawn from the Becker and Tom's framework of rise and fall of families. So in this, according to this Becker and Tom's framework, when financial markets are complete, then the parents' investments in children will not depend on parental income because parents can always borrow against the future earnings of the children. But when financial markets are incomplete, then parental investments in children will depend upon the parental income. So when health shocks affect the households, then the resources are diverted to medical expenditure or the resources available are reduced itself. Hence, the financial resources devoted to schooling might reduce, but there might be other channels through which the health shocks can affect investments in children. For example, parental involvement in children's education and caregiving might reduce. Children's time itself might be devoted more to household and market production activities as the opportunity cost of children increases and there might also be psychological effects. So this diagram shows the different pathways through which investments in children can be affected. Now, coming to the empirical evidence, it is not possible to identify the specific pathways in an empirical framework. So the focus has always been on the cumulative effect on the children's educational attainment. Also, there are different measures of human capital accumulation used in the literature and these capture input, output and outcome indicators. Much of the empirical work has concentrated on Africa. This is because after the spread of AIDS epidemic, millions of children were orphaned in Africa. So the studies have investigated if there are significant differences in schooling between the orphans and the non-orphans. Now, most of the work have found that paternal health shocks, especially maternal deaths, affect children's enrollment in school and also reduce their completed years of schooling. But there are very few studies that have actually done the effect of parental health shocks on children's investments in schooling outside the countries where there is no epidemic done. And also there are some estimation challenges that has to be taken into account which I'll be detailing later. So in my study, I used the data from Young Life's Project which aims to study childhood poverty over a period of 15 years through household and child surveys. Now, this study, this project has been conducted in four countries and in India it is conducted in the state of Andhra Pradesh. Now, two cohorts, age groups of children are being followed. One is the Younger Cohort of 2011 children which were born in 2012. That is, they were one year old in 2012. And then in 2002, and then there are a group of 1,008 children which were eight years old in 2002. Three rounds of the survey have been completed and in my analysis I'm using all the three rounds. The attrition rate compared to other panels survey is very low at 3.6%. Now, in this study, particular study, I'm only using the Young Life's children are included in the analysis, the schooling outcome of other children in the household are not studied. This is mainly for two reasons. Young Life's is not a random sample of all the households in a particular community but it is a random sample of households with a one-year-old or an 80-year-old in a particular community. Also, important factors like child health and learning ability are available only for the Young Life's children. Coming to a profile of the Younger Cohort, this is a profile of the Younger Cohort. These children were one-year-old in 2002, that is in round one, and they were eight-year-old in round three. So in the round three, when they were eight-year-old, 99.2% of the children were enrolled in primary or pre-primary education. But you see that the minimum age of this cohort in the particular academic year is 6.95 years, that is around seven years, so when they are expected to be in grade two. But you find that 6.5% children were not enrolled or they were still enrolled in pre-primary and 12% were attending grade ones, so they were lagging behind. So is there a temporary delay in the initiation into primary school? For this, I use two outcome variables. The first variable is an indicator variable which takes value one. The child is enrolled in grade two and above or zero and it takes zero otherwise. The second is a continuous variable defined as follows. So this particular variable takes value one if the child has completed the grade appropriate for the age and it takes values more than one if the grade completed is higher than that expected of the age and vice versa. This is a profile of the grade, age-specific grade attained by the children by the parental health status. Coming to the older cohort, the older cohort where one year old, sorry, the older cohort were eight years old in the round one and they were 15 year old in the round three. In round one, 97% of them were enrolled in a primary school in R1. But when they moved to the round three, they only 75% of them were enrolled in school. So is there a termination of schooling due to parental health shocks? For this, again, I construct two outcome variables. One is an indicator variable that takes value one if the child continues to go to school and zero otherwise. But dropping out of school might not necessarily mean that there is lower educational attainment. If the child is able to go back to school after once the household recovers from shock. So I'm using another variable in terms of grades advance that is grade completed in R3 minus grade completed in R1. This is again a profile of the grade advancement by parental health status. Now, moving to the methodology, I use a conditional logic procedure here. Mainly for two reasons. Conditional logic procedure compass only those children from the same community. That is it retains communities where there are both enrolled and dropout children. And also it controls for community level factors like access to schools and health centers which might influence the children in a community. In the case of continuous variables, I use least square regression analysis with community fixed ethics. The main variable of my interest is self reported parental health shocks. These are defined as a serious illness or death of a father or a mother of a young lifestyle between the rounds. Now, the other independent variables may be classified into three categories. The child characteristics, household characteristics and school characteristics. The child characteristics include the age, gender, birth order and number of siblings. In the case of household characteristics, you have initial years of schooling of, sorry, includes years of schooling of mother and father, initial wealth quartile groups of the children, whether the household belongs to socially disadvantaged communities. And in the school characteristics, they are controlled for quality of nearest primary schooling. Before I go and estimate my model, I find that there might be factors that might bias the estimates. For example, there might be unobserved time invariant factors. Health shocks are itself not random events because households facing health shocks may display certain characteristics like social status and mobility that might also determine school attainment. So failure to control for these characteristics will generate bias estimate. Also, there might be unobserved time varying factors. For example, there might be incidents that happened during this period which have affected both the parent and health status as well as the schooling attainment of the children. Examples of this include the local weather shocks like droughts and floods, parental job loss, child morbidity, et cetera. So I take into account unobserved time varying factors by controlling for any other income shocks faced by the households. And I also take into account changes in child's health status and so on. In the case of the first unobserved, to check for the endogeneity issues, I perform two empirical tests. Firstly, I check whether health shocks are persistent. That is, health shocks are correlated over time using a dynamic panel model. So here I find that the lab health shock does not predict the present health shocks, which shows that the health shocks are not persistent over time. Because if they are persistent over time, there might be some unobserved characteristics that might be driving my results. Now, I also check if the children with low school participation are also more likely to have parents that face health shocks. That is, if lab non-school participants' participation predicts the future health shocks of the parents. And I also find this is not the case. So proceed to estimate the model. Here, the main results I find that the parental health shocks, in the case of younger cohort, reduces the age-specific enrollment as well as the grade attainment. Now, in many studies, which have not controlled for quality of primary schooling as well as the ability of the child, for example, here I have controlled for the initial learning abilities of the child. You find that the children who have low initial cognitive abilities are also the more, are also the ones who are more likely to drop out. If this is not controlled for, then this might give an over, you will give an over. The estimates might be upward biased. So I control for these factors as well, which are not done in the literature before. So the main conclusions of my study are higher the years of schooling. In the case of younger children, I find that there is a temporary delay in the enrollment in the primary education, while in the case of the older cohort, the schooling attainment is reduced by 0.26 years due to parental health shocks. I also find, which I've not shown here, in the early childhood, maternal shocks are more important than the paternal shocks, than paternal shocks, which mainly affects the child's human capital development to time devoted to childcare. But in the later stage, because for the older children, the opportunity costs are higher than for the younger children. So paternal shocks become more important because loss of income drives the children to be dropped from school and they are sent to work. So, and also other income shocks like child like job loss and parental job loss and child's initial cognitive ability are significant predictors of schooling attainment of children. The other factors are the mother's education and father's education in a significant contribute to the schooling attainment of children. In the case of older cohort, I found that the dropout rates are higher among the older children, among female children, and those who have larger number of siblings. Similarly, wealthier households have, children belonging to wealthier households have more likely to continue into education. And also migration of the household from a particular community into a different community, that negatively impacts the child's education. Now, I've also performed various robustness checks to sample selection, that is I conditioned on both the parents, Halai and Arun. This is just to see that the initial, the health shocks that might have occurred before Arun should not be having later on effects. The effect should not be continuing into later. And I also conditioned on no migration from the community. This is because most of the migration that I've observed from in my data set is distressed migration that is searching for a job outside their community and so on. So they might be affected, they might be affected by other income shocks as well. So I also conditioned on no migration. Then I use different indicators of child health in my analysis. For example, I use changes in BMI, changes in height for age, weight for age, Z scores. I also use indicators of self-reported health shocks, like health shocks reported by the parents to the children. For example, if the children face any serious illness or injury during the period and so on. And I also, in the literature, we find that the credit access to credit market is a very important factor that affects the investment of children's human capital. So I also use indicators for borrowing constraints faced by the households, like access to formal and informal credit markets, whether they are able to borrow, do they borrow at a very high interest rates or do they borrow from the money lenders and so on. Coming to the implications of my study, I find that in my earlier study, I found that households that are low on socioeconomic status are also more vulnerable to health shocks. Therefore, this in turn reduces the future economic well-being of the children through graduate school participation and therefore it perpetuates the poverty and inequality. Now, the state of Andhra Pradesh has a high enrollment rate of 100.76 in the primary education, but this drops to 79.12 in the upper primary education. So maybe policy interventions to retain children in school can be explored by the state. Also, there are studies that have proved that conditional cash transfer programs like Progress have actually retained children in school and they are able to mitigate the income shocks, income shocks faced by the households. So thank you.