 This is a joint paper with Prajiv and Shaheen. It is on impact of income and non-income shocks on child labor. The motivation for these studies is that in countries like Tanzania where poverty is high, say for example, one-third is below the national poverty line two-third of the population is below the 1.25 dollar poverty line. So poverty is a big issue. At the same time these people also face shocks and unfortunately their coping mechanisms are very limited due to various factors. In fact the literature point to several of such factors. Poverty itself is an issue. There are other issues like imperfect credit markets, imperfect land and labor markets, access to assets. All these decide what kind of coping mechanism or at what level they use. So the basic objective of this paper is to examine the relationship between household income and non-income shocks and child labor and to see if the availability of other coping mechanisms or strategies such as social protection mechanisms, access to credit and asset holdings reduce child labor. We rely on the literature and try to see whether consumption is smoothing and the availability of other coping arrangements can explain this relationship. So the economy we are talking about is an economy where the parents make all the decisions including that of the child sending him to school or enabling him to join the labor market. In this particular model there is one parent, one child and the parent decides on the child's human capital development to go to school or asking him to work. So household derives utility. Usual consumption says type utility function. Plus we modify a little bit to incorporate a human capital portion of it taking the literature on the human capital aspects. Parents participate in the labor market and derive income F. So basically they supply labor but also it depends on the shocks as well as household characteristics like, for example, level of education of the parent. So the household problem is to maximize the utility function subject to its budget constraint. Now here the child also generates certain level of income. The time allocated is 1 minus whatever the time spent on education in the labor market. And the solution for this maximization problem is given in 5 where the child labor is dependent on the parent's income, the shocks, the characteristics of the household and also level of time spent on the education and also the random variable. Now the second aspect we want to look at is how the household behaves if the household has access to assets. So the budget constraint is slightly different where now the household has financial assets from which he derives certain level of income. So the first order condition also changed a little bit to reflect that as in 7. The third model we want to look at is now the household has both access to asset as well as access to credit. Again the budget constraint will reflect the previous variable as well as the last two components where now out of the income certain portion is paid at interest on the borrowings he has made in the period. The solution is given in model 9. So basically what we are trying to do is to use Tanzanian data to estimate models reflected in 5, 7 and 9. The data we are trying to use is the latest rounds of Tanzanian National Panel Survey which was initially conducted in 2009 and the second round was in 2011. In these two rounds we notice that the attrition is very low because the over 90% of the round one households were re-interviewed to make sure that they are in the system in the second round. We have a fairly large number of households but for the particular estimation we use roughly about 4,000 children meaning roughly about 700 households. Our outcome measures are children's round two work patterns whether they work in the labour market as child labour and then their pattern of human capital development and a measure of food security taking into account the questions raised. If you look at the human capital component we are restricting it to ages between 7 and 15 because they are the ones who go to the school and we are focusing on them. So on the controls and buffering mechanisms at the child level we limit controls to age and gender, girls and boys. At the household level we use parental education and household size. As for buffer mechanisms we use access to credit we use whether the household has bank account as a proxy and also durable goods as proxy for collateral assets which enable them to borrow from banks. The primary measure of household income shock is crop shocks. It can be due to many reasons, drought, flooding, pest issues and all of that. The data provides because there are particular questions asked whether they have lost their crops during a particular period. And also another shock, a non-income shock we use whether there was a death in the immediate family especially within the household and whether it had an impact on child labour related outcomes. So in terms of empirical strategy we are interested in investigating the relation between child labour intensity and measures of parental income, crop shocks and credit constraints. We have couple of challenges. Potential simultaneity of child labour and parental income and the second one is permitted variable bias with respect to crop shops and child labour. So to address these two challenges we use, we have adapted four-fold strategy to address some of these concerns. We use parents' level of education as a proxy for parental income to avoid potential simultaneity issues include a broad range of control variables including household control such as the size of household and size of household accessibility to land holdings. We investigate whether the household like because the shocks are correlated with household child or parental characteristics and also use regional fixed effects. We use also household fixed effects as a probabilistic check. Now on the question of whether the shocks are exogenous with respect to child labour and labour and whether they are transitory we use a linear probability model to regress crop shocks against having individual, parental and household control variables. But the results in general in credence to causal interpretation of the effects of crop shocks in round one, subsequent outcomes but in one particular case especially whether it is transitory or permanent one we see the round one shocks somewhat related to shocks in round two as well but we see the magnitude is very small so we assume that it is not a serious problem. I am not going to show you all the results because of the limitation of time but particular variable that I am interested in shown in each of the estimation. Now first one the outcome is total child labour hours the variable I am interested in crop shock in round one and see how the child behaves in round two. You will see in all these estimators we use parallel OLS estimation. We will see that in the full sample it is highly significant 7.6 hours increase in child labour in round two but it is mainly coming from male or boys. Now the first one in the full sample in fact the 7.6 is roughly 12% increase of child labour with respect to the mean sample mean and also it is about 15% with respect to the boys so it is quite significant. In the second one outcome is child agricultural labour hours. See again we see that crop shocks increase child labour in the full sample as well as in the male sample as well again quite similar to the previous one. Here also the percentage impact is very high especially in the agricultural sector especially in the boys one it is about 42% with respect to the sample average. Child wage labour hours again we will see it is not only in the full sample but also both boys and girls the impact is somewhat significant but the magnitude is not as high as in the other two areas. In this particular case what it means is that when there is crop shocks they are moving from wage labour to agriculture looking at the previous one and the other two. And the other one we use is household labour hours. Again here it is quite significant especially with respect to male. And also now if you look at the next one when you use the outcome of whether the child has left the school because of a shock you will see that the main effect is significant 3.9% increase in people living out but it is mainly coming from girls 5.6 and 5 to 6% increase in but if you take the sample average it is quite high in fact it is about 35% increase in terms of percentage living out of school. So the basic message coming out of these set of investigations is that households are not fully able to cope with the agricultural shocks they face by just in children's level of labour hours meaning that despite working additional hours children's food security continues to be an issue because the last one in particular is about question of whether the household had food security issue. You will see that both boys and girls suffer from hunger despite working in the labour market as children. So overall results suggest that crop shocks lead not only to an increase in child labour hours but to a change in the compression child time use. Children are spending less time engaged in wage work but more time in agriculture work and girls are less likely to be in school. We also consider buffering effects as I said in the model 7 and 9. So the first one you use is the bank account access to credit. You will see that access to credit significantly reduces the child labour hours. A similar outcome is a completely different outcome in the household hours. Well of course this could be due to the fact that the proxy variable we are using may have some limitations. On the food security one also you will see a reduction in food security issues for hunger if people have access to credit. Another one as I mentioned is the assets. We get very similar results as in the access to credit where when people have assets they can mitigate some of the effects. So leaving school is very much similar when people have assets. Negative impact on the adverse effects of schooling. Similar results on food security as well. So two points. Significant buffer effects that go in the direction we expected. Access to bank account seems to buffer children against hunger when household experience agricultural shocks. So I will leave this and go to policy implications. Four points we come across is significant of income shocks on child labour and the resulting impact on future human capital development. Possible mitigating measures as indicated by some buffer effects especially when you have access to assets and credit. Possibilities of using household characteristics such as parental education which is not shown but have some positive impact on reducing child labour. Possible adverse gender biases of some coping strategies. Girls suffering heavily in the face of household income shocks. So these are some of the issues that probably we have to address in terms of policy. Thank you. Good. Thank you very much.